The role and influence of external review mechanisms
on corporate green bonds
Master’s thesis Finance
Presented to the Faculty of Economics and Business
University of Groningen
In partial fulfilment of the requirements for the degree of Master in Finance
By
Remko Hofman
1. Introduction
The issues of sustainability and corporate social responsibility are increasingly relevant in the world of finance. In the past decades there has been a boom in both corporate activities as well as academic research related to such issues. This is observed for example by the rapid increase in investor demand for socially responsible mutual funds (USSIF report, 2018). One market which came into existence due to these developments and has recently experienced exponential growth, is the market for green corporate bonds. This market started in 2007 when the European Investment Bank issued the first green bond which was then called a ‘climate awareness bond’. Since then, the market has rapidly grown. Green corporate bonds are a way for corporations to issue debt and simultaneously signal to the market their commitment to their environmental performance. This relatively new market is fast-growing but nevertheless still ill-defined. A green bond, similar to other fixed income securities, is issued by a corporation to raise capital for projects, assets, or other investments. In addition, green bonds have the goal to actively contribute to Environmental, Social, and Governance (ESG) related policies. However, there is no universal definition of what that should entail. In most cases, the use of proceeds of green bonds should be CO2 reduction or neutral projects,
or other environmentally friendly investments. However, the lack of definition and regulation also paves the way for greenwashing. Firms could try to use the emerging market of green bonds by making a bond appear green yet in reality only use it as a way to appear socially responsible. One way in which corporations can signal their commitment to their ESG goals and mitigate suspicions of greenwashing is by having their bond classified as green by an external party. But since there is no universal definition of a green bond or a globally
accepted regulations system, there is not one global standard for how bonds are classified as green. Thus, different options arise which might all have different advantages and
disadvantages. And while academic research of the green bond market and empiric work on announcement effects is increasingly being published (Tang and Zhang, 2020; Bagnoli and Watts, 2020), little to no research exists on this choice of classification mechanism by the firms. Baker, Bergstresser, Serafeim, and Wurgler (2018) do find that certain pricing and ownership effects are strongest for bonds that are externally certified as green, but they do not explore the motives behind this certification, nor do they distinguish between different
We examine the yield spread of the bond and at the liquidity of the stocks of the corporation. We consider 3 forms of external classification: third-party assurance, second-party opinion, and green bond rating. The third-party assurance and green bond rating consist of a review of the green bond’s framework with the Green Bond Principles (GBP) by audit firms or rating agencies, respectively. The second-party opinion not only reviews the green bond’s
framework but also the eligible projects or investments and is done by ESG service providers or scientific experts. We rank these different review mechanisms in terms of the scrutiny of the review and the reliability of the review with respect to the signal the issuer is trying to send to the market. The second-party opinion is defined as the most extensive and reliable review mechanism, and the third-party assurance is considered the least extensive and reliable of the three. By ranking the mechanisms on reliability, we not only consider motives for choosing an external review, but we can also investigate the motives for choosing a specific review mechanism.
We find that profitability, size, leverage, and market-to-book ratios are important characteristics in deciding upon the classification mechanism. Larger and highly levered firms have a tendency to pick the more reliable and thorough review mechanisms, whereas smaller and more profitable corporations opt for less extensive channels. We argue that large corporations do so in order to mitigate potential greenwashing fears and to send a signal to the market about their commitment to their environmental goals. Furthermore, firms with emission reduction targets and initiatives are positively associated with more reliable review channels. This finding further suggests that firms genuinely wanting to improve their
environmental performance signal their commitment by choosing an extensive review mechanism. Contrarily, we find that corporations with high carbon emissions are less likely to pick the more reliable, scrutinous review methods. This is evidence suggesting that greenwashing is actually occurring in the green bond market and thus the market is justified in having its concerns about greenwashing. Therefore, we expect the presence of an external review mechanism to have positive effects on the terms of the bond. As suspicions for greenwashing are justified, and the presence of an external review mechanism can be used to mitigate such suspicions, we expect green bonds with an external review mechanism to be rewarded by the market with lower yields. Similarly, as companies send a more reliable signal to the market if they let their green bond be externally certified, we expect the external review to have a positive association with the stock liquidity of the issuer.
coefficients are negative as expected but are not significant at any level. It has to be noted that yield data for the green bonds was limited, and thus this low number of observations causes the power of the statistical tests to be low. The evidence in favor of the hypothesis that a review mechanism is associated with higher stock liquidity of the issuer of a green bond is more pronounced but still inconclusive. Only for one of the three review mechanisms studied is the liquidity significantly positively associated with the review compared to the situation without external review.The remaining two mechanisms are positively associated with stock liquidity but are not statistically significant. We do find strong evidence of a positive relation between the reliability of the review and the liquidity of the issuer’s stock. On average, the more reliable the review mechanism, the higher the stock liquidity. This is strong evidence in favor of the hypothesis that the degree of reliability of the review mechanism is of significant importance. This evidence suggests that issuers picking a more reliable review mechanism benefit the existing shareholders.
The remainder of the paper is structured as follows. Section 2 provides a review of the literature and the hypothesis development. Section 3 describes the sample construction and the methodology used. Section 4 presents the results of the regression analysis. Finally, section 5 concludes.
2. Literature review and hypothesis development
Although the market for green bonds is rapidly increasing (Climate Bonds Initiative, 2016; 2017), academic research is still picking up. The research so far has mostly examined one topic: the pricing of green bonds, and the associated effect of good corporate social performance on bond yields. The general consensus seems to be that green bonds are issued at a negative premium compared to conventional bonds. This negative premium is found for corporate green bonds by Baker et al. (2018), and Zerbib (2017; 2019). Is his 2019 work, Zerbib also finds that the green bond premium is more pronounced for financial and low-rated bonds. Ehlers and Packer (2017) also find a negative green bond premium at issuance but note that their performance in the secondary market is similar to that of conventional bonds. Oikonomou, Brooks, and Pavelin (2014) find that good corporate social performance is rewarded with lower yield spreads, and conversely bad social performance is penalized with higher spreads. This result is supported by Ge and Liu (2015). One opposing result comes from Karpf and Mandel (2017). They study US municipal green bonds in comparison with conventional bonds and find that green municipal bonds are penalized with lower prices and higher yields.
bonds. However, green bonds from private issuers actually have a positive premium with respect to their conventional counterparts. This positive premium for private issuers can be mitigated if they commit to let their green bonds be verified as green by a third party. This is one of the very few studies where the certification mechanism of a green bond is actively studied. Here, the authors distinguish between external certification and no certification, and conclude that external certification serves as a credible commitment to a green project which causes the bond the be issued at more favourable terms. However, within the external
certification group no distinction is made between the different forms of external
certification. One other study that considers the certification mechanisms of green bonds is Ehlers and Packer (2017). They provide an overview of the historical certification
mechanisms and how they have evolved. They then argue that the classification mechanisms should become more consistent across the globe, and that with more consistent green bond standard the market can be better developed further.
In this paper, we will start by examining the drivers of choosing a review mechanism. We consider both financial and environmental firm factors which might influence this
decision. Effectively no previous literature exists in which the drivers for this choice have been investigated. Bachelet et al. (2019) do make a distinction between institutional and private issuers of green bonds and argue that private issuers use external reviews as a way of reducing information asymmetry and mitigating concerns of greenwashing. However, no specific firm characteristics are investigated. Our sample consists only of private issuers, meaning all firms in this sample have a motivation to have their bond externally reviewed thereby reducing greenwashing concerns. Most greenwashing concerns will arise with large, well-known firms, or firms which would have the most to gain from a greener image such as high-emission firms. As these issuers would attract the most attention upon their green bond announcement, they will be most prone to greenwashing concerns and thus will have the highest motivation to mitigate those concerns. Also, firms with emission reduction targets and initiatives will have a strong motivation to have their bond certified as green to signal their commitments to these goals. Concretely, we hypothesize that:
Hypothesis 1a: Firm size is positively associated with the likelihood of choosing a review mechanism
and:
Hypothesis 1b: Total carbon emissions and the presence of emission reduction targets and initiatives are positively associated with choosing a review mechanism
external review mechanism is on the financial results of the bond issuance. Based on the previous literature, we expect the presence of a review mechanism to have beneficial effects on the green bond yields. The evidence in this literature suggests the existence of a green bond premium: green bonds are rewarded by the market with lower yields. Therefore, we expect the presence of a review mechanism to benefit the yields, as it adds reliability to the issuance and mitigates possible concerns about greenwashing. Thus, we specify our
hypothesis as:
Hypothesis 2a: The presence of an external review mechanism in green bond issuances is associated with lower bond yields.
One other topic that is being studied increasingly in recent years is the effect of green bond issuance on the existing shareholders of the issuer. Empirical studies on the stock price reaction of green bond issuance announcements are increasingly being published in recent years. One such study by Baulkaran (2019) finds significant positive cumulative abnormal returns post announcement. It also shows that firm growth is positively related with the CAR, which is consistent with the view that investors see green bonds as value-enhancing. Another extensive empirical study on green bond issuance and its consequences is performed by Tang and Zhang (2020). Using an event study methodology, they find a positive relationship between green bond announcement and stock price reaction. This reaction is more profound for first-time issuers. They do not find strong evidence of a green bond premium; green bonds do have lower yield spreads than conventional bonds across a broad sample but when
comparing green bond yield spread to same-company conventional bond spreads, no significant difference is found. Furthermore, they find that institutional ownership, stock turnover, and stock liquidity are all significantly higher after green bond issuance. Overall, they conclude that green bond issuance is beneficial to the existing shareholders of the issuer.
We build on this literature by investigating the effects of the presence of an external review mechanism in green bond issuances on stock liquidity of the issuer. As this review mechanism increases the reliability of the issuer’s commitment to their environmental performance, we expect to corroborate the findings by Tang and Zhang (2020) and find a positive relationship between the review mechanisms and the stock liquidity. We specify this hypothesis as:
Hypothesis 2b: The presence of an external review in green bond issuances is associated with higher stock liquidity of the issuer.
third-party assurance report. This report, typically provided by accounting or audit firms, states whether the bond is aligned with a reputable international framework for green bonds, such as the Green Bond Principles or Green Loan Principles. The second mechanism is the Second party opinion. Here, an ESG service provider (such as Sustainalytics or Vigeo Elris) or scientific expert (such as CICERO or CECEP) assesses the greenness of the bond
framework, and the eligible projects or assets. Some might also provide the bond framework with a sustainability rating to give a qualitative indication of the greenness of the planned use of proceeds. The third option is a green bond rating. This is when a rating agency such as Moody’s or S&P assesses the bond’s alignment with the Green Bond Principles and the integrity of its green credentials. In terms of reliability, we make the following distinctions: the second-party opinion is seen as the most reliable review mechanism. Firstly, because the external party is either an ESG service provider specialised in assessing exactly such
questions as how green a bond is exactly, or scientific experts in the field. Secondly, because not only the bond framework is assessed, but also the greenness of eligible project or assets to which the proceeds of the green bond may be allocated. Next in line in terms of reliability is the green bond rating. In this review, both the alignment of the bond framework with the GBP and the integrity of the green credentials of the bond are assessed. This places it above the third-party assurance where only the alignment with the GBP is checked. Based on this distinction, we can further investigate the effects of the different review mechanisms on the bond yield and stock liquidity. We hypothesize the following:
Hypothesis 3a: A higher degree of reliability in the review mechanisms is beneficial for the green bond yields
and:
Hypothesis 3b: The reliability of the review mechanisms is positively associated with the stock liquidity of the issuer
The argumentation is a continuation of the reliability argumentation: the more reliable a review mechanism is, the stronger the signal to the market that the issuer is serious in its socially responsible efforts. This again mitigates the possible suspicions of greenwashing and attracts more investors, which is why the bond will be rewarded by the market.
distinguish between the reliability of different review mechanisms and examine the degree to which that reliability has an effect on bond yields and stock liquidity.
3. Data and methodology
3.1. Sample construction
Data on corporate green bonds has been obtained from 2 sources: the Climate Bond Initiative (CBI) and Bloomberg. Both provide extensive data of bond issuers, projects, use of proceeds, and the review mechanism used. The CBI dataset covers green bond issuances from 2007 until 2019, the Bloomberg dataset covers the period from 2010 until 2019.
CBI employs a 3-step process to determine whether a corporate bond can be classified as green. First, bonds with a green theme or label are identified. Then, the projects or assets behind those issuances are evaluated on their compatibility with the Climate Bonds
Taxonomy1. Finally, the allocation of proceeds to those compatible projects and assets is
assessed. CBI bases their assessment on publicly available data such as corporate reporting, bond prospectus, issuer website, rating agencies, and media coverage.
Bloomberg bases their classification of a green bond on the classification made by the issuer. If the issuer either labels their own bond as green or identifies the bond as
environmentally sustainable oriented, Bloomberg will classify the bond as green. The issuer has to have clear statements about their commitment to allocate the proceeds of the bond to projects, assets, or activities that fall under the Green Bond Principles (GBP) use of proceeds categories2. However, they do not require any other reporting on the use of proceeds
management to label bonds as green.
We find a significant overlap in the coverage of green bonds between the two datasets. Around 1/3rd of the specific bond issuances is detected in both datasets. The
remainder of the issuances comes from the more comprehensive CBI dataset. The final sample includes data for each green bond issuance on the amount issued, country of issuance, industry, and the classification mechanism. For each issuance, the chosen classification mechanism, if any, is included. After the initial search of the CBI and Bloomberg green bonds database, a sample of 6120 individual green bond issuances is formed. For our research purposes, we only consider publicly listed corporations. This reduces the sample to 5387 issuances. Next, firm-specific data on environmental preparedness factors and CO2 emissions
for green bond issuers is collected from public reports. Using machine-learning algorithms, information on internal processes of green bond issuers relating to their emission-abatement strategies is collected. This information is gathered from the corporate climate impact reports and other climate risks mentioned in the annual reports. All available reports are scanned by
1
More details about CBI’s green bond taxonomy are detailed in their website.
2
the algorithm on the following keywords, including their variants and combinations: climate, emissions, abatement, energy, strategy, oversight, risk, processes. This results in information about emission reduction targets and strategies, managerial climate action incentives and strategies, and board oversight. A full overview of the exact data gathered on environmental preparedness of the corporations is found in appendix A. The environmental preparedness data from each issuer used is from the year before bond issuance to check whether certain preparedness factors influence the certification mechanism chosen. After matching this sample of firms for which environmental preparedness data and co2 emission data is available with the data of green bond issuances, 158 green bond issuances of 78 individual
corporations remain. However, not for all 158 issuances all data is available. Because we took the environmental preparedness data from the year before issuance and the data is collected from multiple sources, some variables have more limited data than others. In our sample, for 122 of the 158 bond issuances the data on environmental preparedness is
complete. Most variables have between 130 and 150 observations. Panel A of table 1 shows the sample of 158 bonds by country and industry. The main industries in this sample are Financial Services and Energy, followed by Construction and Transportation. Other
industries include Electrical Equipment, Forest and Paper Products, and Telecommunications. Most bonds were issued in Europe (Spain, France, UK) and the US. Panel B shows the year of issuance and the amount issued of the bonds. We see that from 2014 onward an increasing number of green bonds are issued, the total amount being doubled in both 2018 and 2019. Most bonds are issued at less than $250 mln, however most come from the last 3 years. The amount at which the bonds are issued range from a low of $1,32 mln to a high of over $3,4 bln. The average bond in the whole sample was issued at $532 mln.
Financial data on green bond issuers from this new sample is collected from various sources, including ORBIS and Thomson Reuters Eikon.
To match the dataset with the data from ORBIS, for each individual firm a unique ORBIS firm identifier number had to be found. This was done manually by searching in ORBIS on firm name and country. For each firm, the unique ORBIS id was noted once it was made certain the right firm had been selected. Once all ORBIS ids had been found, the
financial data was then retrieved from ORBIS. This includes data on the return on assets, total assets, leverage, and market-to-book ratios. Again, this data is not complete for the 158 issuances. Return on assets, total assets, and market-to-book ratio are rather complete with 145, 145, and 136 observations, respectively. Due to data limitations, we were not able to construct the leverage measure for banks. This results in the leverage measure being available for 101 of the 158 issuances.
Historic daily return and volume data is collected from Eikon. This data is used to construct a measure of illiquidity later on. Yield data is available from the Bloomberg dataset
Table 1: Green bond characteristics.
and is converted to yield spread using the website of the US government treasury3. Similar to the environmental preparedness data, all financial data (except yield spread) is collected from the year before the green bond issuance. The liquidity measure is constructed for all publicly traded companies, which results in the measure being available for 143 out of the 158 bond issuances. Yield data is obtained from Bloomberg but is limited. Only for 52 out of the 158 bond issuances is the yield available. Table 2 provides summary statistics of the firm
3
https://www.treasury.gov/resource-center/data-chart-center/interest-
rates/Pages/TextView.aspx?data=yield
Panel A: Green bonds by country and industry
Industry
Country Transport Construction Energy Financial Services Others Total
Australia 0 0 0 1 0 1 Belgium 0 0 0 1 0 1 Brazil 0 0 1 0 4 6 Canada 0 0 1 3 0 4 Chile 0 0 0 0 1 1 China 3 0 0 0 1 4 Finland 0 0 0 0 1 1 France 0 0 7 17 0 24 Germany 0 0 3 1 0 4 Ireland 0 0 1 0 0 1 Japan 5 4 0 4 7 20 Netherlands 3 0 0 6 1 10 Poland 0 0 0 1 0 1 South Korea 1 0 0 0 3 4 Spain 0 9 17 6 2 34 6 Sweden 0 2 1 1 2 Taiwan 0 0 0 0 3 3 United Kingdom 0 0 7 3 2 12 United States 0 1 7 9 4 21 Total 12 16 45 53 32 158
Panel B: Green bonds by year and amount issued
Issue year Amount issued (in USD, millions)
< 250 250-500 501-750 1000 751- > 1000 Total issued (in USD, millions) Average dollar amount
2013 0 1 0 0 0 1 500 2014 1 4 0 0 2 7 849 2015 1 2 5 2 0 10 528 2016 3 0 2 3 2 10 657 2017 2 5 4 2 4 17 689 2018 11 6 5 5 6 33 601 2019 30 6 11 6 9 62 448 2020 10 1 4 2 1 18 338 Total 58 25 31 20 24 158 532
Table 2: Firm characteristics.
Variable Mean Median Std. Dev.
Return on assets 2.45 1.92 3.13
Total assets (USD, bln) 447.46 79.62 756.56
Debt to equity 10.59 4.18 47.48
Stocks to total assets 0.05 0.03 0.06
This table reports the summary statistics for firm characteristics such as profitability, size, and leverage. Reported values are from the year before issuance for each bond issue.
characteristics of issuers. It includes measures for profitability, size, and leverage. The return on assets, total assets, and debt-to-equity averages are sensitive to outliers so the median will give a better estimation of the ‘average’ firm. The median firm has a return on assets of 1,92%, with total assets of almost 80 billion US dollars. Total assets within the sample vary enormously as indicated by the high standard deviation. The smallest firm has total assets of around 400 million USD whereas the largest firm has assets worth over 2,6 trillion USD. The median D/E ratio is 4,18.
3.2. Methodology
The methodology used is a regression analysis of 2 stages. In the first stage, we estimate each individual classification mechanism with the set of firm characteristics as described in table 2, a set of country dummies for all countries included in the sample, a set of industry dummies for all included industries, and the set of environmental preparedness data. We also construct a categorical variable which takes the value 0 if no review
mechanism is chosen, 1 if the review mechanism is third-party assurance, 2 if the review method is green bond rating, and 3 if the review chosen is second party opinion. This
categorical variable will be used to distinguish the different review mechanisms on reliability and examine the effects of the degree of reliability. The categorical variable increases with the reliability of the review mechanism chosen. A positive relation between it and a variable can be interpreted as a higher associated degree of reliability of the review mechanism with an increase of that variable. The firm controls include return on assets as a proxy for
profitability, total assets as a proxy for firm size, debt-to-equity as a proxy for leverage, and stocks-to-total assets as a proxy for market-to-book ratio. The environmental
preparedness data includes questions on whether the firm (prior to the green bond issuance) provided managerial incentives for climate change issues, if there is board overview, if climate change issues are incorporated in the business strategy, if there were active emission reduction targets or initiatives, if the goods or services sold are sold at low-carbon emissions, the amount of operational spending on energy, and total yearly CO2 emissions. For each
regarding environmental awareness which might influence firms in deciding upon if or which review mechanism to choose. Think of the difference between developed and developing countries, or the commitment of different countries to the Paris climate change pledge in 2015. We control for industry as that can also be of influence in this decision. Firms in the (non-durable) energy or transport sector might have different motivations than banks or other financial services firms. As we are interested in the influence of firm-specific factors, we add the country and industry controls. The regression thus looks as follows:
𝐸𝑥_𝑟𝑒𝑣𝑖 = 𝛼1+ 𝛽1𝑃𝑖 + 𝛽2𝑆𝑖+ 𝛽3𝐿𝑖 + 𝛽4𝑀𝑖+ 𝛽5𝐼𝑖 + 𝛽6𝑂𝑖+ 𝛽7𝐵𝑖+ 𝛽8𝐸𝑖 + 𝛽9𝑇𝑖 + 𝛽10𝐽𝑖+ 𝛽11𝐺𝑖+ 𝛽12𝑁𝑖 + 𝜀𝑖 (1)
where Ex_revit is the chosen review mechanism, Pit is profitability, Sit is size, Lit is leverage,
Mit is market-to-book ratio, Iit is managerial incentive, Oit is board overview, Bit is
climate-change integrated business strategy, Eit is carbon emissions in metric tonnes, Tit is emission
reduction target, Jit is emission reduction initiatives, Git is low carbon goods or services, and
Nit is operational spending on energy. For readability purposes, all individual country and
industry dummies is not included in this or the following regression formulas, but they were included in running the regressions.
In the second stage, we estimate the financial results of the bonds using 2 dependent variables: the yield spread of the bond and a measure for stock liquidity. We estimate the yield spread with the same set of independent variables and controls but now also with the classification mechanism included, again controlling for country and industry. We thus obtain a set of 4 yield spread regressions which investigate whether the classification mechanism has an impact of the yield of the green bonds. The regression looks as follows:
𝑌𝑖𝑒𝑙𝑑 𝑠𝑝𝑟𝑒𝑎𝑑𝑖 = 𝛼1+ 𝛽1𝑃𝑖 + 𝛽2𝑆𝑖 + 𝛽3𝐿𝑖 + 𝛽4𝑀𝑖 + 𝛽5𝐼𝑖+ 𝛽6𝑂𝑖 + 𝛽7𝐵𝑖 + 𝛽8𝐸𝑖 + 𝛽9𝑇𝑖+ 𝛽10𝐽𝑖 + 𝛽11𝐺𝑖 + 𝛽12𝑁𝑖 + 𝛽13 𝐸𝑥_𝑟𝑒𝑣𝑖 + 𝜖𝑖 (2)
where all variables are defined as before.
Because the liquidity measures are highly skewed, we take the natural logarithm to obtain a normal distribution. This illiquidity measure is a log-transferred measure, and the
interpretation is the higher the measure, the higher the illiquidity and thus the less liquid the stock. This measure is a high-frequency daily measure. For each green bond issuance, we take the arithmetic average starting from the day before issuance and going back one year. This measure is then regressed the same way the yield spread is, namely:
𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦𝑖 = 𝛼1+ 𝛽1𝑃𝑖+ 𝛽2𝑆𝑖 + 𝛽3𝐿𝑖+ 𝛽4𝑀𝑖 + 𝛽5𝐼𝑖 + 𝛽6𝑂𝑖+ 𝛽7𝐵𝑖 +𝛽8𝐸𝑖 +
𝛽9𝑇𝑖 + 𝛽10𝐽𝑖 + 𝛽11𝐺𝑖 + 𝛽12𝑁𝑖 + 𝛽13 𝐸𝑥_𝑟𝑒𝑣𝑖+ 𝜖𝑖 (4)
where all variables are again defined as before, and we control for country and industry.
4. Results
4.1. Review mechanism
Table 3 shows the results of the first set of regressions, examining which firm characteristics and preparedness factors are decisive for corporations when choosing their review mechanism. Due to the data limitations described earlier, and the large number of variables in the regressions, the final sample for this set of regressions is 72 observations. This is the sample for which all financial and environmental preparedness data is available.
We see that return on assets has no significant impact on the third-party assurance mechanism, a positive significant impact on green bond rating, and a negative highly
Table 3: Green bond review mechanisms.
Review mechanism
Third-party
assurance Green bond rating Second-party opinion Categorical
Return on assets 0 .03** -.035*** -.046*** (.003) (.014) (.012) (.017) Total assets -.124*** .113*** .22*** .762*** (.022) (.032) (.032) (.106) Leverage -.01** .014* .032*** .113*** (.005) (.008) (.009) (.019) Market-to-book -4.919*** 8.379*** -11.943*** -23.992*** (1.014) (2.815) (2.219) (5.094) Management incentive .205** -.306* .084 -.155 (.082) (.163) (.153) (.27) Board oversight -.014 .111 -.138* -.206 (.038) (.093) (.081) (.185) Carbon emissions .011 .004 -.104*** -.293*** (.011) (.024) (.027) (.053)
Emission reduction target .064 -.179 .875*** 2.332***
(.126) (.282) (.291) (.445)
Emission reduction initiative -3.918* 5.215 15.144*** 51.942***
(2.166) (3.856) (4.434) (9.422)
Operational energy spending .006 .015 .005 .05
(.008) (.019) (.022) (.043)
_cons 2.777*** -2.969*** -2.499** -10.659***
(.536) (.834) (.959) (2.625)
Observations 72 72 72 72
R-squared .925 .866 .948 .963
The table shows the regression results of the firm and environmental controls on the review mechanism. Total assets and carbon emissions have been log-transformed for readability purposes. The 3 review mechanism variables are dummies. All environmental controls are dummies except for carbon emissions and operational spending, which are numerical and categorical, respectively. Heteroscedasticity-consistent robust standard errors are in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
like a small increase for tripling your total assets but given the fact that just the firms in our sample range from total assets of approximately $400 million to $2,64 trillion, a 22%
increase is rather significant. The leverage effects are less significant, economically speaking. An increase in the debt-to-equity ratio of 1 is associated with a 3,2% increase in the likeliness of choosing the second-party opinion. This suggests that a firm with an equal amount of equity but double the debt of a comparable firm is only 3,2% more likely to choose the second party-opinion mechanism. The significant negative relationship with third-party assurance is surprising, as that would suggest larger firms are more likely to choose no review over a third-party assurance. The categorical variable suggests that with an increase in firm size also comes an increase in the degree of reliability of the preferred review
mechanism. The larger, more leveraged corporations will prefer more reliability in their review mechanism, while the smaller, generally more profitable firms will do the opposite. By choosing the most reliable review mechanism of their green bond, large firms want to send a signal that their green bond indeed will be used for environmentally friendly projects or assets.
is in line with our expectations. Carbon emissions seem to have no significant effect on third-party assurance and green bond rating individually, however when moving to the second-party opinion and the categorical variable we observe a strong negative relation. This implies that the higher the carbon emissions of a firm, the less likely it is to choose the second-party opinion. High-emission firms also have a negative relation between carbon emissions and the reliability of the mechanism chosen. This is evidence opposing hypothesis 1b. We argued that high emission firms would be more likely to opt for a review mechanism as their green bond issuances would attract more attention and this is the way to mitigate greenwashing concerns. The evidence suggests that these greenwashing concerns are justifiable as high emission firms seem to be less inclined to let their green bonds be reviewed. The opposite is true for emission reduction targets and initiatives. Firms with emission reduction targets and
initiatives are actually more likely to go for the second-party opinion mechanism and have a highly significant positive relation with the reliability of the mechanisms. This evidence supports hypothesis 1b. These results imply that when firms have emission reduction targets or initiatives, and are thus serious in their environmental efforts, they actually do try to send a strong signal to the market by choosing the second-party opinion. For the other preparedness factors, some significant coefficients are found, for example for management incentive and board oversight, which suggest a higher likeliness for the less reliable mechanisms, but no conclusive evidence is found. The factors of climate-related business strategy and low-carbon goods or services are omitted in the regressions because of collinearity.
To test these results for robustness, we ran our models on the subsample where only countries and industries with more than 1 bond issuance were included. The results are shown in appendix B1 and have the same interpretations as the full sample. One difference is that in the subsample, the environmental preparedness factor of whether a firm provides low-carbon goods or services is also included and highly significant. This factor has positive relations with second-party opinion and the categorical variable, both being significant at the 1% level. This evidence is in line with the argumentation that firms choose more reliable review
4.2. Yield spread
In the second stage, we first investigate the yield spread for the different review channels. Table 4 panel A reports the estimations on firm characteristics, environmental preparedness factors, and the review mechanisms. We can instantly see that the limited data availability of the yield spreads causes some difficulties. Firstly, because of the low number of observations, the R2 is extremely high for this set of regressions (namely 1 for all), and all firm characteristics and environmental preparedness factors are highly significant. Secondly, the third-party assurance variable is omitted by Stata because it is collinear with the leverage variable. This combination of a low number of observations and a high number of
independent variables means the results do not provide conclusive or representative evidence. Firm profitability, size, and leverage all seem to result in higher yield spreads. Contrarily, all environmental factors, excluding board oversight, seem to be rewarded with lower yield. The factors not reported are omitted due to collinearity. We find that the coefficients of the second-party opinion and the categorical variable are negative as expected by hypotheses 2a and 3a. However, they are not statistically significant. The positive coefficient of the green bond rating would suggest evidence against hypothesis 2a, but it is also not significant. This is a curious result, given the fact that all other variables are highly significant. We also run the same model for the subsample where only countries and industries with multiple bond issuances are included. These results are shown in appendix B2. In this subsample, we actually do find significant relationships between the review mechanisms and the yield spread. The green bond rating increases the yield spreads whereas the second party opinion decreases it. This is conflicting evidence for hypothesis 2a, where we would expect the presence of any external review to lower the yield spread. Finally, a negative relation is found between the reliability of the review and the yield spread. This is evidence in support of hypothesis 3a. According to this model, the more reliable the external review the lower the yield spread. However, in this subsample the number of observations is again very low. So, although these results provide some evidence both supporting and opposing hypotheses 2a, and supporting 3a, the econometric power of this evidence is low.
Table 4: Green bond yield spread.
Panel A: Yield spread estimations including leverage and environmental preparedness factors
Yield spread
(1) (2) (3) (4)
Third-party assurance
Green bond rating .395
(.177) Second-party opinion -.388 (.174) Categorical -.375 (.168) Return on assets .176*** .17*** .168*** .163*** (.022) (.019) (.018) (.016) Total assets 6.047** 5.407*** 5.379*** 5.325*** (1.171) (.884) (.872) (.848) Leverage .047** .045*** .044*** .043*** (.008) (.008) (.007) (.006) Market-to-book 15.477 (6.918) Management incentive -15.494*** -14.388*** -14.172*** -13.763*** (2.482) (1.987) (1.891) (1.708) Board oversight 3.981*** 3.997*** 3.975*** 3.932*** (.562) (.569) (.56) (.541) Carbon emissions -.658*** -.562*** -.593*** -.652*** (.104) (.061) (.075) (.101)
Operational energy spending -1.036*** -1.042*** -1.036*** -1.025***
(.156) (.158) (.155) (.15)
_cons -130.088** -116.427*** -114.983*** -112.246***
(25.774) (19.667) (19.022) (17.798)
Observations 23 23 23 23
R-squared .999 .999 .999 .999
Panel B: Yield spread estimations excluding leverage and environmental preparedness factors
Yield spread
(1) (2) (3) (4)
Third-party assurance -.347
(.532)
Green bond rating -.123
(.64) Second-party opinion -.25 (.303) Categorical -.142 (.127) Return on assets .004 .005 -.001 .006 (.025) (.028) (.027) (.028) Total assets -.175 -.218 -.108 -.12 (.22) (.161) (.21) (.228) Market-to-book .45 2.098 -.115 -.019 (5.985) (5.966) (5.596) (5.834) _cons 4.545 4.537 2.319 2.701 (5.715) (3.639) (4.837) (5.196) Observations 45 45 45 45 R-squared .933 .933 .934 .936
We first note that the number of observations has doubled in this model. In this model, all the review mechanisms have a negative sign as we would expect. But again, the coefficients are not significant at any level, thus providing no evidence for hypotheses 2a and 3a. One stark contrast between the two full models is the significance of the firm characteristics. In the first model, all coefficients are positive and highly significant. In the second model the
coefficients are either negative or positive but very close to 0, and all insignificant. Considering the evidence from all 3 models, we conclude that the presence of a review mechanism does not significantly improve the yield spread of a green bond. Neither is the degree of reliability of the review mechanism significant in improving yield spreads.
4.3. Liquidity
We also investigate the effect of the different review mechanisms on the stock liquidity of the issuers. These results are presented in table 5. It should be noted here again that the measure used is a measure of illiquidity, the higher the measure the less liquid the stock. Starting with the firm characteristics, firm size seems to be the biggest factor in stock liquidity, which intuitively makes sense. The larger and better known a corporation is, the more liquid its stock will be as its stock is well known by investors and will be traded more often, especially if recent performances have been either very good or bad. This is why the negative relations between total assets and the illiquidity measures is expected, as it indicated that stock become more liquid as total assets grow. Conversely, highly levered corporations seem to have lower stock liquidity. The literature on capital structure has suggested a causal relation between liquidity and leverage. As equity financing becomes more expensive, debt financing rises (Acharya and Viswanathan, 2011). Frieder and Martell (2006) argue that both the relation is bi-directional; liquidity affects leverage, but the capital structure also
significantly impacts liquidity. Therefore, using the same logic from Acharya and
Viswanathan (2011), this adverse relationship is expected. When leverage increases it implies that equity financing is expensive at that moment, which is indicated by a lower stock
Table 5: Green bond issuer stock liquidity.
Stock liquidity
(1) (2) (3) (4)
Third-party assurance -1.7
(2.31)
Green bond rating -1.419***
(.488) Second-party opinion -.443 (.97) Categorical -1.039*** (.166) Return on assets .076* .111*** .059 .03 (.042) (.037) (.051) (.03) Total assets -1.051*** -.629*** -.737*** -.068 (.37) (.12) (.225) (.105) Leverage .037 .066*** .067 .174*** (.038) (.024) (.041) (.032) Market-to-book 16.946 38.658*** 20.188 -.184 (15.347) (9.708) (13.01) (5.441) Management incentive -2.06** -1.579* -2.223** -3.057*** (.958) (.787) (.944) (.789) Board oversight -.032 -.203 -.111 -.087 (.844) (.722) (.803) (.747) Carbon emissions .31** .397*** .257 -.052 (.127) (.122) (.202) (.134)
Emission reduction target .816 2.211** 1.301 2.453**
(1.227) (.977) (.91) (.94)
Emission reduction initiative .275
(2.825)
Low-carbon goods or services 9.751 20.119* 22.686 71.534***
(17.519) (11.789) (19.654) (15.527)
Operational energy spending -.124 -.017 -.121 -.12
(.127) (.106) (.111) (.087)
_cons -6.703 -33.302** -26.849 -85.128***
(25.282) (13.502) (20.287) (16.387)
Observations 70 70 70 70
R-squared .982 .984 .982 .987
The table presents the regression results of the review mechanism, firm characteristics, and environmental controls on the stock liquidity. The stock liquidity measure is defined as the log of the absolute daily return over the dollar volume, where for each issuance the average of the year from before issuance is taken. The higher the measure, the less liquid the stock. Heteroscedasticity-consistent robust standard errors are in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
Overall, the existence of a review mechanism might improve liquidity depending on the mechanism chosen. Our evidence for hypothesis 3b is much stronger, however. As the
the subsample where countries and industries with only 1 bond issuance are excluded (appendix B3).
5. Conclusions
In this paper, we have investigated the review mechanisms of green bond issuances. This review mechanism is a signal of the issuer to the market about its dedication to its environmental performance. This can be an accounting or audit firm checking the green bond’s framework for legitimacy, or a thorough review of the bond by ESG service providers or scientific experts. More specifically, we investigate which firm characteristics are driving the choice in classification method and whether certain environmental preparedness factors influence this decision. Additionally, we examine the effects of the different review
mechanisms on bond yields and stock liquidity and hypothesise a positive association for both. Furthermore, we expect this positive association to increase with the reliability of the chosen review mechanism.
We find that profitability, size, and leverage are the key characteristics that influence the choice of review mechanism. Larger, more levered firms on average are more likely to let their bond be externally reviewed. In addition, we also find a strong positive relation between firm size and leverage and the degree of reliability of the chosen review mechanism. Smaller, typically more profitable firms are less likely to choose a review mechanism. They also have a strong negative relation with the degree of reliability of the review mechanisms. We show that corporations with high carbon emissions are less likely to choose an extern review and additionally are less likely to opt for a more thorough review. Conversely, corporations with active emission reduction targets and initiatives are more likely to choose an extern review and are strongly positively associated with the reliability of the review. These findings suggest that firms which are genuine about their environmental commitments actively use external review mechanisms as a way to reduce information asymmetry and mitigate greenwashing concerns. However, it also suggests that those greenwashing concerns are justified, as high-emission firms avoid review mechanisms, especially the more scrutinous ones.
We find no convincing evidence that the existence of a review mechanism benefits the issuers in terms of lower green bond yields. Neither is evidence found which suggests that the reliability of the review is rewarded by the market with lower yields. These results are limited by the small sample size, making statistical inference less powerful. After
investigating the effects of the review mechanisms on the stock liquidity, we find weak but inconclusive evidence in favour of the hypothesis that the existence of a review mechanism in green bond issuances improves stock liquidity of the issuer. Only for one of the three
Overall, we find that external review mechanisms are used as signals to the market about the issuer’s environmental commitment. However, these signals do not seem to be rewarded with better bond terms in the market. For existing shareholders, the presence of a review
References
Acharya, V., & Viswanathan, S. X., 2011. Leverage, moral hazard, and liquidity. The Journal of Finance, 66(1), 99-138.
Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series effects. Journal of financial markets, 5(1), 31-56.
Bachelet, M. J., Becchetti, L., & Manfredonia, S., 2019. The green bonds premium puzzle: the role of issuer characteristics and third-party verification. Sustainability, 11(4), 1098.
Bagnoli, M. E., & Watts, S., 2020. On the corporate use of green bonds. Journal of Economics & Management Strategy, 29(1), 187-209.
Baker, M. P., Bergstresser, D. B., Serafeim, G., and Wurgler, J. A., 2018. Financing the response to climate change: the pricing and ownership of u.s. green bonds. Working paper. (No. w25194). National Bureau of Economic Research.
Baulkaran, V., 2019. Stock market reaction to green bond issuance. Journal of Asset Management, 20(5), 331-340.
Climate Bonds Initiative, 2016. Bonds and climate change: the state of the market in 2016. Climate Bonds Initiative, 2017. Green bonds highlights 2016.
Ehlers, T., & Packer, F., 2017. Green bond finance and certification. BIS Quarterly Review September.
Fong, K. Y., Holden, C. W., & Trzcinka, C. A., 2017. What are the best liquidity proxies for global research? Review of Finance, 21(4), 1355-1401.
Frieder, L., & Martell, R., 2006. On capital structure and the liquidity of a firm's stock. Working paper. Krannert School of Management.
Ge, W., & Liu, M., 2015. Corporate social responsibility and the cost of corporate bonds. Journal of Accounting and Public Policy, 34(6), 597-624.
Karpf, A., & Mandel, A., 2017. Does it pay to be green? Working paper. Université Panthéon-Sorbonne, Paris.
Lesmond, D. A., 2005. Liquidity of emerging markets. Journal of financial economics, 77(2), 411-452. Oikonomou, I., Brooks, C., & Pavelin, S., 2014. The effects of corporate social performance on the
cost of corporate debt and credit ratings. Financial Review, 49(1), 49-75.
Tang, D. Y., & Zhang, Y., 2020. Do shareholders benefit from green bonds? Journal of Corporate Finance, 61, 101427.
USSIF, 2018. Trends in Socially Responsible Investing.
Zerbib, O. D., 2017. The green bond premium. Working paper. University of Tilburg.
Appendix A: Environmental preparedness data.
Description
in main text Full question Variable type
Management
incentive CC1.2 - Do you provide incentives for the management of climate change issues, including the attainment of targets? Dummy Board
oversight Board Oversight Dummy
Business
strategy C3.1_Are climate-related issues integrated into your business strategy? Dummy Carbon
emissions CC8.2. Please provide your gross global Scope 1 emissions figures in metric tonnes CO2e Numerical Emission
reduction target
CC3.1. Did you have an emissions reduction or renewable energy consumption or production target that was active (ongoing or reached completion) in the
reporting year? Dummy
Emission reduction initiative
CC3.3. Did you have emissions reduction initiatives that were active within the reporting year (this can include those in the planning and/or implementation
phases)? Dummy
Low-carbon goods or
services CC3.2. 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? Dummy Operational
energy
Appendix B1: subsample where industry and country contain more than 1 issuances, review mechanism
Review mechanism
Third-party
assurance Green bond rating Second-party opinion Categorical
Return on assets 0 .03** -.035*** -.046*** (.003) (.014) (.012) (.017) Total assets -.124*** .113*** .22*** .762*** (.021) (.031) (.032) (.105) Leverage -.01** .014* .032*** .113*** (.005) (.008) (.009) (.018) Market-to-book -4.919*** 8.379*** -11.943*** -23.992*** (1) (2.776) (2.188) (5.023) Management incentive .205** -.306* .084 -.155 (.081) (.16) (.151) (.266) Board oversight -.014 .111 -.138* -.206 (.037) (.092) (.08) (.183) Carbon emissions .011 .004 -.104*** -.293*** (.01) (.023) (.027) (.052)
Emission reduction target .064 -.179 .875*** 2.332***
(.124) (.278) (.287) (.438)
Emission reduction initiative .073 -.089 -.98** -3.046***
(.203) (.426) (.427) (.646)
Low-carbon goods or services -3.991* 5.304 16.124*** 54.988***
(2.173) (3.884) (4.407) (9.277)
Operational energy spending .006 .015 .005 .05
(.008) (.019) (.021) (.042)
_cons 6.769*** -8.274* -18.623*** -65.647***
(2.404) (4.172) (4.318) (9.006)
Observations 70 70 70 70
R-squared .925 .865 .948 .963
Appendix B2: subsample where industry and country contain more than 1 issuances, yield spread
Yield spread
(1) (2) (3) (4)
Third-party assurance
Green bond rating .989**
(.274) Second-party opinion -.96** (.266) Categorical -.974** (.258) Return on assets .047 .354*** .345*** .332*** (.036) (.049) (.047) (.043) Total assets 3.843*** 7.757*** 7.641*** 7.505*** (.185) (1.271) (1.239) (1.188) Leverage -.005 .12*** .116*** .13*** (.015) (.02) (.019) (.019) Management incentive -23.95** -.021 .235 .684 (6.262) (.376) (.448) (.576) Board oversight 2.752*** 5.868*** 5.777*** 5.65*** (.087) (.875) (.85) (.807) Carbon emissions -2.28* 2.019** 1.892** 1.667*** (.829) (.363) (.328) (.265)
Low-carbon goods or services 9.277***
(.911)
Operational energy spending -.72*** -1.525*** -1.502*** -1.47***
(.014) (.238) (.231) (.22)
_cons -25.235 -204.363*** -199.097*** -200.371***
(15.559) (34.135) (32.674) (31.065)
Observations 23 23 23 23
R-squared .999 .999 .999 .999
Appendix B3: subsample where industry and country contain more than 1 issuances, liquidity
Stock liquidity
(1) (2) (3) (4)
Third-party assurance -1.7
(2.293)
Green bond rating -1.419***
(.485) Second-party opinion -.443 (.963) Categorical -1.039*** (.165) Return on assets .076* .111*** .059 .03 (.042) (.036) (.051) (.03) Total assets -1.051*** -.629*** -.737*** -.068 (.367) (.119) (.224) (.104) Leverage .037 .066*** .067 .174*** (.038) (.024) (.04) (.032) Market-to-book 16.946 38.658*** 20.188 -.184 (15.237) (9.638) (12.916) (5.402) Management incentive -2.06** -1.579* -2.223** -3.057*** (.951) (.781) (.937) (.784) Board oversight -.032 -.203 -.111 -.087 (.838) (.717) (.797) (.742) Carbon emissions .31** .397*** .257 -.052 (.127) (.121) (.201) (.133)
Emission reduction target .816 2.211** 1.301 2.453**
(1.219) (.97) (.903) (.933)
Emission reduction initiative .275
(2.805)
Low-carbon goods or services 9.751 20.119* 22.686 71.534***
(17.393) (11.705) (19.513) (15.415)
Operational energy spending -.124 -.017 -.121 -.12
(.126) (.105) (.111) (.086)
_cons -6.703 -33.302** -26.849 -85.128***
(25.101) (13.405) (20.142) (16.269)
Observations 69 69 69 69
R-squared .982 .984 .982 .987
