• No results found

Determinants of the Issuance of Corporate Green Bonds and the Impact on Environmental Performance

N/A
N/A
Protected

Academic year: 2021

Share "Determinants of the Issuance of Corporate Green Bonds and the Impact on Environmental Performance"

Copied!
32
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

1

Determinants of the Issuance of Corporate Green Bonds

and the Impact on Environmental Performance

Thesis MSc Finance

Focus area: Sustainable Society Elles Nijmeijer

S2666146

University of Groningen Faculty of Economics and Business

Supervisor: Dr. S. Homroy 10 January 2021

Abstract

This thesis examines how corporate strategies determine the issuance of corporate green bonds and their effect on CO2 emissions. I execute a cross-section analysis for the determinants and a fixed effects test for the effect on CO2 emissions. I find that companies already engaging in environmental sustainability activities, such as implementation in their business strategy and producing low-carbon products, are more likely to issue green bonds. This supports my signalling hypothesis. However, I find no evidence that the issuance of green bonds does directly reduce CO2 emissions. This supports my greenwashing hypothesis and signalling hypothesis. Companies issue green bonds to signal their environmental engagement to the market rather than to reduce CO2 emissions in the short run.

(2)

2

1.Introduction

Since the first issuance of the green bond in 2007 labelled as the Climate Awareness Bond (CAB)1, the market for green bonds has increased significantly. Green bonds are a financial instrument that requires that an issuing organization invests or reinvests in climate and environmental projects to, for example, reduce GHG emissions. Therefore, green bonds can contribute to the goals of the Paris Agreement which contains the limitation of global warming to below 2 degrees Celsius.2 To commit to this goal, the European Commission President Ursula von der Leyen has set a target to raise green bonds, which is 30 percent out of 750 billion euros issued in bonds under the Next Generation EU program. With the introduction of this target, it has become clear that green bonds have won popularity and importance from a governmental and policy point of view. Green bonds are in this respect used as a policy instrument to create support among investors and corporate organizations for the climate issues and the urge to tackle this problem. This development is reflected in the rapid increase of green bonds issued to a total value of $259 billion dollars in 2019 according to the Climate Bond Initiative (CBI). Although green bonds, particularly corporate green bonds, are increasingly popular, there is still little knowledge on which type of firms are more likely to issue green bonds. Financial gains do not seem to be a strong motivation for firms to issue green bonds. Past research has found mixed results for a financial effect after the issuance of green bonds. Where Tang and Zhang (2018) find a short-term positive effect on shareholder value, while Fernando, Sharfman & Uysal (2017) suggest that institutional investors avoid firms with a high environmental score which does not lead to a higher shareholder value. Although Flammer (2020) finds that green bonds announcements do have a positive impact on the stock market, this does not imply major financial benefits regarding financial performance. Therefore, there should be another reason why companies issue green bonds.

There can be two theoretical motivations for corporate green bond issuance. Therefore, I will test two hypotheses. The first motivation is ‘greenwashing’. ‘Greenwashing’ means that companies state that they will invest the loan in green investments, but these projects are not proven to be carbon emission reducing (Trompeter, 2017). To reduce the chance of greenwashing, the International Capital Market Association (ICMA) introduced the Green Bond Principles (GBP) in 2014 (Trompeter, 2017). Nonetheless, this is still a voluntary certification.

The second motivation is signalling. Companies can signal their engagement in environmentally sustainable activities, such as GHG emission and water waste reduction. This

1 This information is retrieved from the European Investment Bank. EIB Climate Awareness Bonds,

https://www.eib.org/en/investor_relations/cab/index.htm

2 More information on the Paris Agreement can be found on:

(3)

3 theory has also been tested by Flammer (2020) who finds support that companies issue green bonds to signal their commitment as the stock market reacts positively to the announcement of the issuance of green bonds. Another way of signalling could be that companies already engage in environmentally sustainable activities but want to show this through the issuance of green bonds and not necessarily primarily focus on the reduction of CO2 emissions. Maltais and Nykvist (2019) find support for this type of signalling where they examine that companies primarily issuing green bonds already have a strategy that engages in environmentally sustainable activities. So, this study gives more insight on the reasoning behind the issuance of green bonds and the effectiveness on CO2 emissions and finds support that companies use green bonds as a signal rather than using green bonds as a useful financial instrument to reduce CO2 emissions. Hence, I find support for the ‘greenwashing’ theory as well.

The data I use for the empirical analysis, is extracted from the Climate Bond Initiative (CBI) and Bloomberg information about green bond issuers. The data on determinant variables and CO2 emissions is retrieved from annual financial and non-financial reports of companies. Firm specific data is extracted from ORBIS. The data contains a full sample of corporate green bond issuers and non-issuers from 2010-2018 from public companies worldwide which report their CO2 emissions.

(4)

4 Second, I examine the impact of issuing green bonds on CO2 emissions in the following years. For this analysis, I use a fixed effects test to examine the change in CO2 emissions after the issuance of green bonds within companies and compare this with a pooled Ordinary Least Squares (POLS) test. I find that there is no statistically significant evidence that the issuance of green bonds decreases CO2 emissions. My findings support the ‘greenwashing’ hypothesis, since I find no evidence that the issuance of green bonds results in a reduction in CO2 emissions. Taken together, my results suggest that companies that already engage in environmentally sustainable activities, like producing low-carbon products or have a climate-issue integrated business strategy, are more likely to issue green bonds, which implies that this does not necessarily improve their environmental performance further. This argumentation is also supported by my findings that there is no evidence that issuing green bonds decreases CO2 emissions in the following years. Therefore, it is not clear whether issuing green bonds does further reduce CO2 emissions, since the companies that are most likely to issue green bonds, already engage in environmentally sustainable activities. These findings both confirm the signalling and greenwashing hypothesis. On the one hand, my findings contribute to the paper of Bagnoli and Watts (2020) who find that companies issue green bonds to cover the costs of low-carbon products. On the other hand, the findings add to Maltais and Nykvist (2019) who state that companies who issue green bonds already engage in environmental policies or targets. This implies that mainly firms that already engage in some sort of climate related activities, issue green bonds. Therefore, this research adds to existing literature of Flammer (2020) by adding another interpretation of signalling climate-issue related activities through the issuance of green bonds. Furthermore, my findings add to Trompeter (2017) who finds that green investment projects are not proven to be carbon emission reducing. However, my findings contradict the research of Flammer (2020) who suggests that certified green bonds reduce CO2 emissions, but it has not become clear whether these certified green bond issuers already engaged in environmental activities in her research. Finally, this research gives more insight in the reasoning behind the issuance of green bonds by corporate firms. This is important, since policymakers can adapt their policies to the findings and focus more on companies that do not issue green bonds yet. Furthermore, future research should focus on other ways of emission reduction from a corporate view so that organizations will work towards a more sustainable world in the future.

(5)

5

2. Literature review

2.1 The development of ESG activities and green bonds

The awareness of climate change and the need to reduce GHG emissions worldwide, led to the introduction of the green bond by the ECB in 2007. However, before the first issuance of the green bond, ESG3 activities and its impact were already examined extensively. Since ESG activities focus, among other things, on environmental sustainability, it pursues partly the same goal as green bonds. However, the reduction of GHG emissions followed by ESG activities is not the topic that has been investigated primarily. Past research majorly focussed on the effects of ESG on financial performance, but results are mixed. Where Xie et al. (2018) find that there is a positive relation between ESG activities and ROA, McWilliams and Siegel (2000) do not find a significant positive relation. However, Xie et al. (2018) do not find evidence of a positive relationship between climate change policies and financial performance. This could imply that climate change policies such as green bonds issuance, does not create a positive impact on financial performance. Besides, Friede, Busch & Bassen (2015) found in their review of over 2000 studies, that the majority suggests a positive relationship between ESG activities and financial performance.

This relationship thus holds for ESG activities and not specifically for green bonds. However, the growth of green bonds on the financial market increases the need for research on this topic specifically. In existing literature on green bonds, the focus is the financial gain of green bonds where Zhou and Cui (2019) find that an announcement of green bonds issuance has a positive effect on a firm’s stock prices, profitability and operational performance which holds for Chinese firms. To build upon this finding, Flammer (2020) suggests a positive effect on the stock market after a green bond issuance by public companies with a worldwide sample. Also Tang and Zhang (2018) find a short-term positive effect of green bond issuance on shareholder value. However, Fernando, Sharfman & Uysal (2017) find evidence that institutional investors avoid firms with high environmental scores, which could imply a negative impact on shareholder value. Taken together, these findings do not imply a clear financial gain of issuing green bonds, so there should be other reasons why companies issue green bonds, which will be discussed in the next sections.

3 ESG stands for Environmental, Social and Governance. This type of activities focuses on

(6)

6 2.2 Determinants of Green bonds

(7)

7 and whether these companies use the issuance of green bonds as a credible signal to signal their environmental engagement to the market. Accordingly, the first hypothesis for this research is:

Hypothesis 1: Companies use green bonds to signal their engagement in environmental sustainability activities to the market

2.3 Green bonds and environmental performance

(8)

8 are effective instruments in fighting climate change and reduce CO2 emissions. Accordingly, the second hypothesis focuses on the impact of issuing corporate green bonds on the CO2 emissions of companies and whether it supports the ‘greenwashing’ theory. Therefore, the second hypothesis that is tested is the following:

Hypothesis 2: CO2 emissions are not reduced as result of the issuance of green bonds.

3. Data

3.1 Data sources Green bonds issuance

The dataset for the green bond issuance is extracted from the Climate Bond Initiative (CBI) and Bloomberg’s reports of green bonds. These are the main sources of information on green bond issuers, the use of proceeds and projects. The dataset contains corporate green bonds issuances for the period of 2007-2019 and the Bloomberg database covers the period of 2010-2019.CBI follows a process to classify a green bond in three steps, which is the following:

1. Identify bonds with a green-themed label

2. Scan the projects or assets for calibration with the Climate Bonds Taxonomy 3. Classify the allocation of proceeds to aligned projects and assets

The assessment of CBI is based on available information sources from issuer websites, bond prospectus, rating agencies, corporate reporting, and media coverage.4

Bloomberg labels a bond as ‘green’ when the issuing company identifies the bond as an environmental sustainability-oriented bond issue with clear additional statements about the commitment to deploy funds towards projects and activities in the Green Bond Principles use of proceeds categories or when the issuing company labels its bond as ‘green’ itself. Bloomberg does not require further qualifications to classify the bond as green.

Next, the information on CO2 emissions and internal processes of green bonds for companies related to their emission-reduction strategies is collected. For this data, machine-learning algorithms are used to extract this information from the annual financial and non-financial reports of companies. Here, I use corporate climate impact reports and climate risks mentioned in the annual reports. All available corporate climate action reports and annual reports are downloaded, and the needed information is extracted through machine-learning algorithms. The training library contains phrases using the following keywords, their variants and combinations: climate, emissions, abatement, energy, strategy, oversight, risk and processes.

(9)

9 Next, the organizations from the CBI file are matched to the organizations in the Emissions files. Matched organizations are labelled as green bond issuers and included in the dataset. The other organizations are labelled as non-green bond issuers. These criteria give a total of 1,746 corporate green bonds, issued by 87 different publicly listed companies, and a total of 6,177 observations for the analysis.

Control group

All organizations that are included in the sample for the extraction of CO2 emissions and internal processes but did not issue green bonds, are included in the control group. This gives a total of 5,570 observations consisting of 2,397 unique publicly listed companies. The CO2 emissions are reported in metric tonnes.

Determinants of green bonds

For the analysis, several determinants for the issuance of green bonds are used. These variables are answers to questions found in annual reports of the corresponding organizations which issued green bonds and organizations from the control group in a period from 2010 until 20185. The following variables are constructed from the survey. Management Incentive is the incentive for the management to engage in climate change issues, including the realization of targets.

Board oversight is the indicator of whether an organization has board oversight to ensure that

executives act in the best interest of its shareholders. Business Strategy is whether climate-related issues are integrated into the business strategy. Emissions Target is whether the organization had an emissions target that was active in the reporting year. Reduction Initiatives means whether organizations had initiatives to reduce emissions that were active in the reporting year. Low Carbon Products is the classification of organizations whether they have any existing goods and/or services classified as low-carbon or if they accredit a third party to avoid GHG emissions. Energy costs is the percentage of the organizations total operational spend in the reporting year on energy.

Emissions data

The emissions data is measured in CO2 emissions. The data is retrieved from the same data source as the determinant variables. In this research, emissions are reported as CO2 emissions in metric tonnes and is converted in natural logarithms. The variable is reported as Log

(Emissions) Firm specific data

In this research, I examine the characteristics of the firms that best predict the issuance of green bonds. The financial data is retrieved from ORBIS. This is done by a manual matching process.

(10)

10 The list of organizations extracted from the machine-learning algorithms, is manually filled in the ORBIS searcher to obtain BvD ID numbers. Subsequently, the BvD ID numbers are used to extract all financial data available in ORBIS. The variables that will be controlled for in this research are the following. Return on Assets (ROA) is the ratio of profit/loss before tax to total assets measured in percentages. Size is the natural logarithm of the total book value of assets and is reported as Log(assets). ROE is the ratio of profit/loss before tax to total equity in percentages. This measures the leverage of the organization. Solvency is the ratio in percentage of the solvency relative to assets. Country data is stated as whether countries meet the goal of the Paris Agreement to keep their emissions under the level of 2 degrees Celsius compatibility. This data is subtracted from the Climate Action Tracker6, where countries are tested whether they are on track with the target considering a country’s fair share effort to the Paris Agreement. In the regression, one country is dropped to avoid perfect multicollinearity.

3.2 Summary statistics Green bonds market

To study the determinants and effects of green bonds, it is necessary to analyse the dataset that I use in this thesis. The descriptive statistics are reported in Table 1. There are in total 3,458 firm-year observations on whether organizations issued green bonds. 150 of these observations are green bonds issuances in the period from 2010 until 2018 for 59 unique public, corporate organizations where some companies issued more than one green bond in the given period. The other observations used in the sample did not issue a green bond, which is presented in column (2) of Table 1 as the control group.

First, I analyse the development of the dependent variable, green bonds, since the first issuance in 2007 with the data from CBI. Figure 1 and Figure 2 present the development of green bonds and shows that from 2013 there was a rapid increase in the number of bonds and the total value issued, which led to an issuance of $259 billion divided over 1804 bonds in 2019. Figure 3 shows that mostly developed economies issue green bonds. In Table 2, the number and value of corporate green bonds issued per country is presented. The countries are divided in developed and emerging economies. The table shows that especially the United States, Sweden, Japan and France issued corporate green bonds in developed economies, while China is the only emerging economy that issues green bonds to the same extent. However, in the sample that I use, the balance is slightly different. Still, the largest portion of green bonds issued is in developed economies, but now Japan, Spain and the United Kingdom are highly represented. This implies that companies in these countries report more information on their internal processes regarding climate-issue related activities. Table 3 reports green bonds issued per industry in the dataset

(11)

11 of my thesis. Industries that mainly issue green bonds are banks and companies that on average have a higher level of GHG emissions like construction, energy and metal smelting companies. Second, I examine the descriptive statistics of the independent variables in the dataset. This is reported in Table 1. Some variable information is not available for all observations every year. They are excluded in the analysis discussed in section 5.

Firstly, I note thatcompanies in the green bonds group have a higher mean on all independent variables regarding the determinants compared to the control group. This implies that companies that issue green bonds, tend to be more involved in environmental policies. This seems logic, since green bonds are part of this policy. However, their CO2 emissions and energy costs are higher which also implies a larger environmental impact. This corresponds with the fact that industries with a higher environmental impact have issued more green bonds compared to other industries. This could mean that for these companies, there is a higher need to engage in environmental policies in order to improve their environmental performance and reduce their energy costs.

Secondly, Business strategy and Reduction Initiatives show perfect multicollinearity. The

Business strategy variable is a dummy and is 1 for all organizations which issued green bonds

in a certain year. For Reduction Initiatives holds that it is 1 for almost all green bond issuers. Besides,Reduction Initiatives and Business Strategy are slightly correlated by 0.32 (Table 6)7. As result, these two variables are omitted in the analysis. However, it seems that all green bond issuers have climate-related issues integrated in their business strategy. This implies that green bonds are part of this strategy. The same holds for Reduction Initiatives. This implies that almost all green bond issuers have initiatives to reduce CO2 emissions. This implies that issuing green bonds are a tool to achieve these reductions.

Thirdly, Green bond issuers have, on average, a higher amount of assets, which implies that larger companies are more likely to issue green bonds. However, the return on their assets (ROA) and equity (ROE) is lower than that of organizations in the control group. For the solvency ratio holds that the control group has a higher mean than the green bonds group. This means that the control group has a higher probability of paying off their long-term debt than Green bond issuers.

Finally, I note that ROE and ROA show a moderately high correlation of 0.49. Therefore, ROE is dropped as control variable in the regression.

(12)

12 Table 1. Summary statistics Corporate Green bonds at the company level

The table provides summary statistics of organizations which provided their CO2 emissions. In column (1), (2) and (3) the

number of observations of the full sample, control group, which did not issue green bonds and the green bonds issuers are provided, respectively. In Column (4), (5) and (6), the mean and standard deviation of the full sample, control group and green bonds group are reported, respectively. In column (7), the Median of the full sample is reported. The table reports the full sample of the data from 2010-2018 of corporate organizations.

N Mean/St.dev Median Full Sample (1) Control (2) Green bond (3) Full Sample (4) Control (5) Green bond (6) Full Sample (7) Management Incentive 2,891 2,601 135 0.691 (0.462) 0.686 (0.464) 0.926 (0.263) 1 Board Oversight 2,891 2,601 135 0.729 (0.444) 0.723 (0.448) 0.911 (0.286) 1 Business Strategy 2,305 2,044 110 0.852 (0.355) 0.845 (0.362) 1 (0) 1 Emissions Target 2,483 2,192 184 0.874 (0.332) 0.879 (0.327) 0.918 (0.274) 1 Reduction Initiatives 2,487 2,192 186 0.952 (0.214) 0.957 (0.204) 0.984 (0.126) 1

Low Carbon Products 2,487 2,192 186 0.731 (0.444) 0.731

(13)

13 0 50000 100000 150000 200000 250000 300000 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

Figure 1. Value of Corporate Green Bond Issuance in

US dollars (in $million)

Total 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019

Figure 2. Number of Corporate Green Bonds Issued

Total 0 100000 200000 300000 400000 500000 600000

Developed economies Emerging economies Multiple countries

(14)

14 Table 2. Corporate green bonds per country

This table presents the total amount of green bonds issued in USD and the number of green bonds issued per country over the period of 2007-2019 where country groups are divided in developed and emerging economies. Also, the number of green bonds in the dataset is presented per country. Amount Issued # of Corporate Green bonds Country USD (million) CBI Dataset Developed economies Australia 11904 40 1 Austria 2584 9 Belgium 705 2 1 Canada 15952 39 8 Denmark 1802 6 France 40617 173 9 Germany 17768 37 6 Italy 12168 24 Japan 12877 95 39 Netherlands 15282 26 4 New Zealand 1843 15 Norway 6996 29 Singapore 4061 18 Spain 20558 44 34 Sweden 19589 235 12 Switzerland 2559 10 United Kingdom 9247 27 31 United States 163846 3758 36 Other countries 10288 30 3 Subtotal 370646 4617 184 Emerging economies Argentina 637 5 Brazil 4367 21 15 Chile 797 5 6 China 79744 222 7 Hong Kong 4802 21 India 6372 23 Malaysia 908 8 South Korea 4277 11 10 Taiwan 1936 18 3 United Arab Emirates 3525 3 Other countries 8866 51 Subtotal 116232 388 41 Total 486878 5005 225

Table 3. Green bonds issued per Industry

All green bonds issued in the period of 2010-2018 included in the dataset per industry.

Industry Number of Bonds Financials Banks 32 Insurance 5 Other 20 Industrials

Construction and engineering 10 Electric utilities and energy traders 10

Chemicals 7

Metal smelting 6

Telecommunication services 5

Building products 4

Forest and paper products 4

Automobiles 3

Consumer products 3

Technology hardware and equipment 3

Power generation 2

Media 1

Others 110

(15)

15

4. Methodology

4.1 Determinants of green bonds

The cross-sectional study examines the determinants of the issuance of corporate green bonds. A cross-sectional test with two methods is used. In the first method, I examine the determinants of green bonds based on data of a one lag period for all independent variables with respect to the issuance year of the corresponding green bond from that organization when this data is available. Otherwise, the observation will be reported as missing value. In the second method, the independent variables will be a one lag variable for the organizations that issued green bonds. For the control group, I report one lag period data of the last available data of the corresponding organization. In this respect, the lagged data could also contain data from two or three periods ago, where one period stands for one year. A cross-sectional test is used to correctly examine the relation to the independent variables in a previous period and the green bond issuance in the successive period. Since not all organizations report the determinant variables each year, the second method of last available data could show different results, since organizations could in fact engage in the activities included in the variables. However, this could bias our results since it could give inaccurate results. Therefore, it is necessary to compare the two methods.

The model that is used for the cross-sectional test is:

(1). 𝐺𝑟𝑒𝑒𝑛𝐵𝑜𝑛𝑑 𝐼𝑠𝑠𝑢𝑎𝑛𝑐𝑒𝑖

= 𝛼𝑖+ 𝛽1𝐷𝑀𝑎𝑛𝑎𝑔𝑒𝑚𝑒𝑛𝑡𝐼𝑛𝑐𝑒𝑛𝑡𝑖𝑣𝑒1𝑖+ 𝛽2𝐷𝐵𝑜𝑎𝑟𝑑𝑂𝑣𝑒𝑟𝑠𝑖𝑔ℎ𝑡𝑖

+ 𝛽3𝐷𝐵𝑢𝑠𝑖𝑛𝑒𝑠𝑠𝑆𝑡𝑟𝑎𝑡𝑒𝑔𝑦𝑖+ 𝛽4𝐷𝐸𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠𝑇𝑎𝑟𝑔𝑒𝑡𝑖+ 𝛽5𝐷𝑅𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛𝐼𝑛𝑖𝑡𝑖𝑎𝑡𝑖𝑣𝑒𝑠𝑖 + 𝛽6𝐷𝐿𝑜𝑤𝐶𝑎𝑟𝑏𝑜𝑛𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑠𝑖+ 𝛽7𝐸𝑛𝑒𝑟𝑔𝑦𝐶𝑜𝑠𝑡𝑖+ 𝛽8𝐸𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑠𝑖+ 𝛾𝑖+ 𝜀𝑖

Where GreenBond Issuance is the dummy variable of the issuance of green bonds, where an organization which issued a green bond is given a 1 where each issuance is included as a separate observation. All other organizations are given a 0. The basis for the dummy variables is the answer to the questions8 subtracted from annual reports and were answered by either ‘yes’ or ‘no’. This is translated in a 1 for the ‘yes’ answers and a 0 for the ‘no’ answers and empty for missing values. The variables are the dummy variable for Management Incentive, the dummy variable for Board Oversight, the dummy variable for Business strategy, the dummy variable for Emissions Target, the dummy variable for Reduction Initiatives and the dummy variable for Low-Carbon Product. The Energy Cost variable is translated to the percentage of the total operational spend. This is divided in percentiles of 5%. This gives 21 percentiles, where 1 is 0% and 21 is more than 95% but less than or equal to 100%. Emissions is the natural logarithm of the CO2 emissions in metric tonnes. 𝛾i represents the control variables and 𝜀i is the residual.

(16)

16 For choosing which model to use for the tests, I compare the Linear Probability Model (henceforth LPM), the Probit and the Logit model which are similar tests for data where the dependent variable is a dummy. I report the Logit model as the baseline, but the LPM and Probit model show similar results.9 Furthermore, I execute a Likelihood Ratio test to examine whether I should include the control variables in the model. This test shows statistically significance at the 1% level. Hence, the control variables should be included to correctly predict the model. Regarding the homoskedasticity of the datasets, this is checked for heteroskedasticity with a White test. The test is statistically significant at the 5% level; hence the data is homoscedastic. However, White standard errors are used to assure the homoskedasticity of the data. However, regarding endogeneity, there might be other variables that predict the issuance of green bonds which are not included in the analysis. This should be considered.

4.2 Effect of green bonds on environmental performance

To examine the effect of the issuance of green bonds on the environmental performance, I use an unbalanced panel data test with the following econometric model for the Pooled OLS regression:

(2). 𝐸𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑖𝑡 = 𝛼𝑖 + 𝛽𝐷𝐼𝑠𝑠𝑢𝑎𝑛𝑐𝑒𝑖𝑡+ 𝛾𝑖𝑡+ 𝑢𝑖𝑡

Where Emissionit are the emissions of a given firm at time t measured in metric tonnes CO2,

DIssuanceit is the dummy of the issuance of green bonds, 𝛾it is the set of control variables and

uit is the error term. A pooled OLS regression assumes that there is no heterogeneity. However,

panel data regresses organizations and green bond issuances over time, so the probability of heterogeneity is high, but is hard to observe. Therefore, I will use a fixed effects model as well for both entity and time fixed effects. The fixed effects regression model is the following:

(3). 𝐸𝑚𝑖𝑠𝑠𝑖𝑜𝑛𝑖𝑡 = 𝛼𝑖 + 𝛽𝐷𝐼𝑠𝑠𝑢𝑎𝑛𝑐𝑒𝑖𝑡+ 𝜇𝑖+ 𝑣𝑖𝑡

Where Emissionit is Emissions for a given firm at time t measured in metric tonnes CO2,

DIssuanceit is the dummy of the issuance of green bonds where all years as from the issuance

year equals 1 and otherwise 0. 𝜇it is the entity fixed effect and vit is the error term.

The model tests for the effects of the issuance of green bonds on CO2 emissions reported by the firm. The reason that I use a fixed effects regression, is that the emissions change over time and I test whether CO2 emissions significantly change when a green bond is issued. Regarding the heteroskedasticity of the dataset, I correct for this by using robust standard errors. However, in this analysis, there might be some form of endogeneity, since CO2 emissions might be affected by other factors than the issuance of green bonds. This might bias the results and should be considered.

(17)

17

5. Results

5.1 Determinants of the issuance of Green bonds

In Table 4, I examine the determinants of the issuance of green bonds through an OLS regression. The results of two regressions are presented. In column 1, I run a test with the independent variables being lagged for 1 period. In column 2, the results of the test with the last available data are presented, which is explained in the methodology. I show similar determinants for the issuance of green bonds for both tests. First, Management incentive shows a significant but negative relation to green bond issuance, but only for the second test with a confidence level of 5%, all other variables being constant. This could be interpreted as when an incentive for the management is present regarding climate change, the likelihood of issuing green bonds, decreases by 0.97 times. However, this seems counterintuitive. One would expect that with management incentives, they would be more likely to engage in environmental policies. Second, Board oversight suggests a negative relation with the issuance of green bonds in column 2 which is statistically significant at the 5% level, ceteris paribus. This implies that the presence of board oversight decreases the likelihood of issuing green bonds by 1.077 times. This could imply that the best interest of shareholders is not per definition the issuance of green bonds, since board oversight makes sure that executives act in the best interest of its shareholders. Third, An Emissions target seems to negatively affect the likelihood of issuing green bonds with a statistical significance of 5%, ceteris paribus, but only for the test in column 1. This suggests that having an emission reduction target decreases the probability of issuing green bonds. An explanation for this could be that when firms already are working on a target, they do not feel the need to issue green bonds to improve their environmental performance. This contradicts Maltais and Nykvist (2019), who found that companies who issue green bonds are already engaging in environmental policies or targets. This also contradicts the signalling hypothesis, since this implies that companies might not use green bonds to signal their emission target. However, the second test does not show a significant result, so the result is not convincing. Moreover, stronger results are shown for the second test regarding Low-Carbon

Products. It shows that when firms have low-carbon products, the likelihood of issuing green

(18)

18 ‘green’ projects through green bonds. There could be several explanations for this. First, it could be that it is not necessary to invest in green projects because they are already a ‘green’ firm. Second, it could be that they invest in ‘green’ projects, but do not issue green bonds specifically. Third, there is a possibility of re-financing existing green projects with green bonds which may not result in an additional reduction of CO2 emissions.

Moreover, the size of the firms which is represented by the natural logarithm of the book value of total assets, has a positive impact on the likelihood of issuing green bonds which is statistically significant at the 1% level for both tests, all other variables being constant. The larger the firm, the higher the probability that an organization is issuing green bonds. Furthermore, the Return on Assets has a slightly negative impact on the likelihood of issuing green bonds which is statistically significant at the 1% level for both tests. However, the economic significance is very small, since the impact is close to zero.

Furthermore, Business Strategy and Reduction Initiatives are omitted from the analysis as result of perfect multicollinearity, which is also explained in the summary statistics section.

5.2 The impact of green bonds on CO2 emissions

In Table 5, the results of the fixed effects model are presented. I execute a fixed effects model to correct for unobserved variables that might influence the results and to test for a relation within firms. In column 1, I regress a Pooled OLS test. In column 2, I regress a Fixed effects model. The effect of issuing green bonds on CO2 emissions is measured. The results from the Pooled OLS for the green bond issuance are all statistically significant. There seems to be a positive relation between the issuance of green bonds and the change in CO2 emissions. Furthermore, the size of the organization positively and significantly affects the amount of CO2 emissions at the 1% confidence level. This could be explained by the fact that larger companies tend to produce more and hence have higher CO2 emissions. Furthermore, the ROA and Solvency ratio show a negative and statistically significant relation at the 1% level. However, this effect is economically negligible. However, the Pooled OLS regression does not correct for heteroskedasticity issues. Therefore, I also run a Fixed effects test.

(19)

19 Table 4. Determining characteristics of corporate firms in relation to the issuance of green bonds This table presents estimates of the cross-section analysis of equation (1). The sample includes public companies which reported their CO2 emissions in the period from 2010-2018. The dependent variable

is a dummy equal to 1 if the company issued a green bond and equal to 0 otherwise. The independent variables are dummies, except for Energy costs, Emissions and the control variables. Standard errors are reported in parentheses. *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively.

Green bond issuance One lag

(1)

(20)

20

5.3. Robustness checks

The regressions need to be checked for their robustness. Therefore, I apply five checks to test whether the outcomes are robust for both regressions, which are shown in Table 8 for the test with 1 lag data, Table 9 for the test with last available data and Table 10 for the fixed effects test, which can be found in appendix 3.

In column 1, instead of regressing with the natural logarithm of Emissions and Total Assets, I execute the analysis with the absolute values. The 1 lag test gives similar results. The latest available data test shows less significant results. The fixed effects model shows significant results when I use absolute values. So, for the 1 lag test and fixed effects test, the results are robust, but the fixed effects test is not robust.

In column 2, I exclude the financial data variables for the OLS regression. For the one lag test, only Board Oversight shows a statistically significant result at the 10% level together with the natural logarithm of Emissions. This is still statistically significant at the 1% level. However, when doing this check for the latest available data, not Board oversight, but Management

incentives, Energy costs, Low-Carbon products and Emissions are statistically significant. This

Table 5. Effect on CO2 emissions as result of the issuance of green bonds

This table presents the estimates of the pooled OLS and fixed effects analysis from equation (2). The dependent variable is the natural logarithm of CO2 emissions in metric tonnes per year. Green Bonds

is a dummy variable where issuance is 1 from the year of issuance onwards and otherwise reported as 0. The other variables are control variables. Standard errors are reported in parentheses. *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively.

CO2 Emissions

Pooled OLS Fixed Effects (1) (2) Green Bonds 1.625** -0.125 (0.712) (0.281) Log Assets 0.150*** 0.016 (0.017) (0.010) ROA -0.020** -0.004 (0.008) (0.003) Solvency -0.024*** -0.001 (0.003) (0.001) Country -0.032*** - (0.003) Constant 9.477*** 9.981*** (0.289) (0.176) Observations 2,066 2,071 R-squared 0.086 0.005 Number of Organizations 961

(21)

21 is different from the benchmark, which did find a significant effect of Board Oversight, but not of Energy costs. The other variables do correspond with the benchmark. The fixed effects test shows similar results.

In column 3, I winsorize the data by 0.5% on both sides to exclude outliers. The results for the one lag period show more significant results, but the latest available data test shows less significant results. Therefore, this check is not robust for both tests. The fixed effects Winsor check shows significant results for green bonds, which is different from the benchmark. In column 4 and 5, I run the tests for the developed countries and developing countries. I divide the countries according to the UN10 report for 2020. Looking at the results for the one lag period data, developing countries show similar results compared to the benchmark, while Developed countries give a significant result for Board oversight instead of Emissions Target. However, for developing countries, the relation between green bond issuance and Low-carbon Products is negative, which contradicts other results. For the latest available data, developed countries show significant results for Board oversight, Low-Carbon Products and Emissions are significant, but for developing countries, Management Incentive, Board Oversight and Energy

Costs are significant. So there seems to be a difference between the developing and developed

countries when it comes to the determinants of issuing green bonds. For the fixed effects model, both developed and developing countries show similar results.

10 Distribution of developing and developed countries is subtracted from The World Economic Situation

and Prospects of the United Nations in 2020 and can be found through the following link.

(22)

22

6. Conclusion

6.1 Discussion and limitations

This research focusses on characteristics of companies’ internal processes that predict the issuance green bonds and the impact of issuing green bonds on CO2 emissions. The determinants that are tested are related to the environmental policy of companies. I analyse public companies that reported their CO2 emissions and execute an analysis for both green bond non-issuers and issuers as the dependent dummy variable for the period 2010-2018 in one dataset. The tests show mixed results. I execute two tests with a different method which do not give similar results for all variables. This indicates that the results are not clear. Only the level of CO2 emissions and the company’s size significantly affect the likelihood that a company issues green bonds for both tests. This implies that an environmental policy does not affect the choice of issuing green bonds. This contradicts the findings of Maltais and Nykvist (2020) who state that companies that issue green bonds already engage in environmental sustainability activities. This also contradicts the signalling hypothesis. Furthermore, there is some evidence that the presence of board oversight negatively affects the likelihood of issuing green bonds. This implies that shareholders do not want companies to issue green bonds. Having an emission reduction target also shows some significant negative relation to the issuing green bonds. This contradicts the findings of Maltais and Nykvist (2020) and implies that companies that have such targets do not feel the need to issue green bonds, while Maltais and Nykvist (2020) find that these types of companies are the main issuers of green bonds. This contradicts the signalling hypothesis, since I find evidence that companies who are engaged in emission reduction do not necessarily signal this through the issuance of green bonds. However, I find evidence that having Low-Carbon products increases the likelihood of issuing green bonds. This adds to the literature of Bagnoli and Watts (2020) who state that companies that produce Low-Carbon Products, are more likely to issue green bonds to cover part of the additional costs.

However, what I find in my dataset is that all corporate green bond issuers already have climate-related issues integrated in their business strategy. This supports the signalling hypothesis and the findings of Maltais and Nykvist (2020).

(23)

23 might be true, since there is no evidence that green bond issuance reduces CO2 emissions in this thesis. However, this type of signalling does not say anything about the effectiveness of these environmental strategies. Wat it does say is that companies are aware of climate change and are willing to put it on their agendas.

Overall, I find supporting evidence for my hypotheses. First, I find evidence for the signalling hypothesis. There is some evidence that companies engage in climate-related activities, like selling Low-carbon products, or having a business strategy with climate-related issues integrated, which increases the likelihood of issuing green bonds. The likelihood of issuance also increases when companies have a higher level of CO2 emissions. This could also imply a signal to the market that they are working on reduction of emissions. However, the evidence is not very strong, since the results are not statistically significant for both tests. Therefore, the signalling hypothesis is not supported in both cases. Second, I find evidence for the second hypothesis that CO2 emissions are not reduced as result of the issuance of green bonds, which supports the theory of greenwashing. This implies that companies use the issuance of green bonds as a signal to the market, rather than issue green bonds to reduce their CO2 emissions. This research has some limitations. First, the sample size of corporate green bond issuers who report their CO2 emissions is small. This makes it harder to correctly predict the results in econometric models. Furthermore, the data that is subtracted from the annual reports of the companies, are written by the companies themselves. This is an arbitrary measure since it is not known whether the data is objective or subjective. So, companies might state that they engage in environmental policies, because this is good for their image, while this might not be the case. Moreover, in the analysis, there are some variables that are automatically omitted due to perfect collinearity while it might determine the issuance of green bonds for a large part and therefore should also be considered in this research. This concerns the climate-issue integrated business strategy and emission reduction initiatives. However, since there are only 110 observations of green bond issuers included in the regression, the sample could be too small to correctly predict the results.

6.2 Future research

(24)
(25)

25

References

Bagnoli, M., Watts, S. G., 2020. On the corporate use of green bonds. Journal of Economics & Management Strategy 29(1), 187-209.

Chiesa, M., Barua, S., 2019. The surge of impact borrowing: the magnitude and determinants of green bond supply and its heterogeneity across markets, Journal of Sustainable Finance & Investment 9(2), 138-161.

Ehlers, T., Packer, F., 2016. Green Bonds, Certification, shades of green and environmental risks, Bank for International Settlements

Fernando, C., Sharfman, M., Uysal, V., 2017. Corporate Environmental Policy and Shareholder Value: Following the Smart Money. Journal of Financial and Quantitative Analysis 52(5), 2023-2051.

Flammer, C., 2020. Green Bonds: effectiveness and implications for public policy. Environmental and Energy Policy and the Economy 1(1), 95-128

Friede, G., Busch, T., Bassen, A., 2015. ESG and financial performance: aggregated evidence from more than 2000 empirical studies. Journal of Sustainable Finance & Investment 5(4), 210-233.

Karpoff, J. M., J.R. Lott Jr., E. W. Wehrly., 2005. The Reputational Penalties for Environmental Violations: Empirical Evidence. Journal of Law and Economics 48, 653–675.

Maltais, A., Nykvist, B., 2020. Understanding the role of green bonds in advancing sustainability. Journal of Sustainable Finance & Investment, 1-20.

McWilliams, A., Siegel, D., 2000. Corporate social responsibility and financial performance: correlation or misspecification?. Strategic Management Journal 21, 603-609

(26)

26 Tang, D.Y., Zhang, Y., 2020. Do shareholders benefit from green bonds?, Journal of Corporate Finance 61

Trompeter, L., 2017. Green is good: how green bonds cultivated into Wall Street's environmental paradox. Sustainable Development Law and Policy Brief, 17(2).

Xie, J., Nozawa, W., Yagi, M., Fujii, H., Managi, S., 2018. Do environmental, social, and governance activities improve corporate financial performance?. Business Strategy and the Environment 28(2), 286-300

(27)

27 Appendices

Appendix 1

Questions for the independent dummy variables as testable determinants of green bonds.

The answers to the following questions are being used to regress the determinants of issuing green bonds. Question 1, 2, 3, 5, 6 and 7 can only be answered with ‘yes’ or ‘no’. Question 4 is answered with a number in metric tonnes. Question 8 is answered with a number from 1-21. Where each number represents a range of 5 percent.

1. Do you provide incentives for the management of climate change issues, including the attainment of targets?

2. Board Oversight

3. Are climate-related issues integrated into your business strategy?

4. Please provide your gross global Scope 1 emissions figures in metric tonnes CO2e 5. Did you have an emissions target that was active in the reporting year?

6. Did you have emissions reduction initiatives that were active within the reporting year? 7. Do you classify any of your existing goods and/or services as low-carbon products or

do they enable a third party to avoid GHG emissions?

(28)

28 Appendix 2

Table 6. Correlations variables

This table shows whether there is a correlation between the variables. Business Strategy and Management Incentives (0.39) and Reduction Initiatives and Emissions Target (0.34) are moderately correlated. Business Strategy and Reduction Initiatives are moderately correlated (0.32) and omitted from the analysis. ROA and ROE are highly correlated (0.46). Therefore, ROE is omitted from the analysis.

Greenbond Management Board Strategy Target Initiatives Energy Products Emissions TotalAssets Solvency ROA ROE Country

(29)

29 Appendix 3

Table 7. Comparison of different cross-sectional regression models

This table presents estimates of the Linear probability model in column (1), the Logit model in column (2) and the Probit model in column (3). The sample includes public companies which reported their CO2

emissions in the period from 2010-2018. The dependent variable is a dummy equal to 1 if the company issued a green bond and equal to 0 otherwise. The independent variables are also dummies, except for Energy costs, Emissions and the control variables. Standard errors are reported in parentheses. *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively.

Green bond issuance

(30)

30 Table 8. Robustness test for Determinants lagged 1 period

This table presents estimates of the robustness checks for the cross-section analysis with 1 lag data. In Column (1), the natural logarithms are excluded. In column (2), the financial data is excluded. In column (3), the data is winsorized at the 0.5 and 99.5 percentile. In column (4) and (5) a distinction between developed and developing countries is made, respectively. The division of countries can be found in table 1.2. The table reports all observations from 2010-2018. Standard errors are reported in parentheses. *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively.

(31)

31 Table 9. Robustness test for Determinants latest available data

This table presents estimates of the robustness checks for the cross-section analysis with last available data. In Column (1), the natural logarithms are excluded. In column (2), the financial data is excluded. In column (3), the data is winsorized at the 0.5 and 99.5 percentile. In column (4) and (5) a distinction between developed and developing countries is made, respectively. The division of countries can be found in Table 1.2. The table reports all observations from 2010-2018. Standard errors are reported in parentheses. *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively.

(32)

32 Table 10. Robustness test Fixed effects model

This table presents estimates of the robustness checks for the pooled OLS and the fixed effects analysis. In Column (1), the natural logarithms are excluded. In column (2), the regression excludes financial data. In column (3), the data is winsorized at the 0.5 and 99.5 percentile. In column (4) and (5) a distinction between developed and developing countries is made, respectively. The division of countries can be found in table 1.2. All firm-year observations from 2010-2018 are included. Standard errors are reported in parentheses. *,**,*** indicates significance at the 10%, 5%, and 1% level, respectively.

CO2 emissions

Excluding logs Excluding financials Winsor Developed countries Developing countries (1) (2) (3) (4) (5) Green Bonds -1115495.69** -0.057 -1112231.12*** -0.040 -0.814 (535,747.81) (0.384) (416,717.776) (0.287) (0.922) Total Assets 0.000 0.000* 0.017 (0.000) (0.000) (0.025) Log Assets 0.015 (0.011) ROA 4,255.410 1,322.367 -0.003 -0.020* (5,265.805) (4,095.862) (0.003) (0.011) Solvency 749.997 998.578 -0.000 -0.003 (2,186.744) (1,701.630) (0.001) (0.003) Constant 1810738.623*** 9.173*** 1445259.527*** 10.456*** 8.415*** (77,520.094) (0.014) (60,628.531) (0.186) (0.474) Observations 2,427 4,009 2,427 1,618 458 R-squared 0.005 0.000 0.009 0.003 0.024 Number of Organizations 1,195 1,953 1,195 729 233

Referenties

GERELATEERDE DOCUMENTEN

When repeating the analysis among only those patients who had not had (online) contact with fellow patients in the past, we found similar results: Theory of Planned

To explore the empirical validity of our typology of university activities in cross-border regions we use OECD study Regions and Innovation: collaborating cross-border in which

The test above was conducted with the yield spread as dependent variable, individual ESG pillar scores as independent variables and the bond-and firm characteristics as

Hypothesis 2b: The presence of an external review in green bond issuances is associated with higher stock liquidity of the issuer.. Delving deeper into the reliability argument,

Since the corporate green bond issuances are expected to yield a positive effect on the environmental outcomes (increase in Env. Score and decrease in CO 2 / Assets) in the year of-

50% best (worst) performing green bond firms compared to their matched conventional peer, in terms of either the ESG, environmental pillar or environmental innovation score. The

However, the negative relationship for mandatory disclosing firms is inconclusive due to the insignificant result between environmental performance and corporate financial

Binne die gr·oter raamwerk van mondelinge letterkunde kan mondelinge prosa as n genre wat baie dinamies realiseer erken word.. bestaan, dinamies bygedra het, en