Greenium, credit rating, and the COVID-19 pandemic

Abstract

We analyze green and conventional bonds during regular market periods and within times of extreme volatility, the COVID-19 pandemic. We find a negative premium (greenium) of 1.6 bp before the outbreak of COVID-19, but during the times of extreme market stress, this greenium widens to 3.5 bp as our results show a significant outperformance of green bonds. The results indicate that green bonds are more resilient during risk-off periods than non-green bonds. In addition, the greenium effect is moderated by the issuer's country environmental performance as the greenium is more pronounced for issuers from non-green countries prior to COVID-19. We do not find differences between green and non-green countries since COVID-19.

Introduction

Corporations have the opportunity to label bonds as green if they use the proceeds of the bonds to fund projects with an environmental benefit, i.e., projects which facilitate a net-zero emissions economy or protect the environment (ICMA 2021). Green bonds can play a significant role in funding climate goals established during the 21st Conference of the Parties (COP) by the United Nations in Paris in 2015 (Weber and Saravade 2019) and facilitating a cleaner production. Understanding the return requirements of green bonds is therefore of considerable importance for the success of a sustainable transformation of the global economy. Numerous studies (e.g., Hachenberg and Schiereck 2018; Zerbib 2019; Dorfleitner et al. 2022; Koziol et al. 2022) observe that investors are willing to accept a yield discount, a “greenium”, when purchasing green bonds. This finding suggests that green bonds are sold to different investors than conventional bonds. To what extent these different groups of investors behave similarly or differently in different market phases has hardly been researched to date.

In this paper, we examine green bond prices in comparison with conventional bonds before and during the COVID-19 pandemic as an unforeseen event for market participants. External shocks, such as the pandemic, initiate structural breaks in market phases and can provide extreme market volatility which leads investors as a consequence to restructure their portfolios to financial instruments, which historically navigated safely through periods of stress (Baur and McDermott 2016; Kinateder et al. 2021). The COVID-19 pandemic which has started in 2020 represented an extreme external shock to international financial markets (Carlsson-Szlezak et al. 2019). The pandemic and subsequent lockdowns in order to ban COVID-19 had severe effects worldwide: Schools and universities closed globally; travelling was hardly possible and public life came to a standstill. Economic effects followed. According to Fernandes (2020), financial markets saw in March 2020 a sharp drop down in equities and stock indices globally registered their biggest one-day falls on record. Industries that suffered the most were oil, gas, and coal (− 50%) as well as aerospace and defense (− 40%).¹ Volatility was at historical highs, levels similar to or above the financial crisis of 2008/2009 (see Ramelli and Wagner 2020; Baker et al. 2020).

Quickly, people feared a loss of income, psychosocial stress, and unemployment (Douglas et al. 2020). Decreases in consumption and global disruption of supply chains affected companies around the globe (Fernandes 2020). Financial markets quickly recovered, but the real economy only saw the first of a number of lockdowns that subsequently followed. How investors and prices of financial instruments, which were developed to facilitate investments that help to protect against climate change, react during times of stress and macroeconomic shocks, is of high interest to academics as well as practitioners. In particular as several experts are warning for decades of severe global risk due to climate change and the subsequent consequences of it. The green bond market only evolved after the recent financial crisis, e.g., green corporate bonds were hardly existent before 2013 (Flammer 2021). Consequently, a number of papers examine the first real stress test for green bonds and the convenience of green bond issues in turbulent times (Yousaf et al. 2021; Arif et al. 2022; Mensi et al. 2022; Naeem et al. 2022; Narayan et al. 2022; Pham and Cepri 2022). Yousaf et al. (2021) as well as Arif et al. (2022) explore the opportunity of using green bonds as a hedge for conventional stock portfolios or various asset classes during the COVID-19 pandemic. Naeem et al. (2022) as well as Narayan et al. (2022) analyze potential diversification benefits of green bonds in times of extreme shocks. On the other hand, Mensi et al. (2022) study the impact of the COVID-19 pandemic on various green bond industries, while Pham and Cepri (2022) look at the spillovers between investor attention and green bond performance under extreme market conditions. We extend previous research and analyze how green bonds react compared to their direct peers, conventional bonds from the same issuer, during the exogenous shock of the COVID-19 pandemic.

As recent surveys by Cortellini and Panetta (2021) and Dervi et al. (2022) underline, the pricing divergences between green and conventional (non-green) bonds are one of the most intensively analyzed areas in empirical studies. Research has discussed the question if green and conventional bonds trade at the same or different levels in many ways (e.g., Ehlers and Packer 2017; Gianfrate and Peri 2019; Bhurjee and Paliwal 2022 for India; Hu et al. 2022 and Chen et al. 2022 for China; Teti et al. 2022 for Italy). Most recent studies conclude that the growing, deeper market of green bonds trades the instruments at a negative premium to conventional bonds (e.g., Hachenberg and Schiereck 2018; Immel et al. 2020; Zerbib 2019; Löffler et al. 2021). Dorfleitner et al. (2023) report higher liquidity for green bonds of corporate issuers with a greenness rating which again indicates divergent investor groups for bonds characterized by different levels of greenness.

Looking at the link between green bond issuances and stock returns, Tang and Zhang (2020) do not find a consistently significant greenium when issuing green bonds but instead a positive reaction of stock prices following bond issuances. On the other hand, Pastor et al. (2022) who analyze green bond and green stock returns do not expect an outperformance of green stocks going forward. They divide returns in ex ante and ex post returns and estimate ex post lower expected returns for green stocks than for brown stocks.

The first aim of this paper is to examine if green corporate bonds as well as financial bonds performed differently during COVID-19 compared to their direct counterparts, conventional bonds. Previous studies on other financial instruments provide evidence that sustainable-linked instruments have performed better during the pandemic than conventional financial instruments (Albuquerque et al. 2020; Pastor and Vorsatz 2020). Albuquerque et al. (2020) analyze stocks during the first quarter of 2020 and provide evidence that stocks with higher environmental and social (ES) ratings generate significantly higher returns, lower return volatility, and stocks held by more ES-oriented investors experience less return volatility. Pastor and Vorsatz (2020) find that US mutual funds with high sustainability ratings performed better during the COVID-19 pandemic, and investors remained focused on sustainability as well as exclusion criteria when reallocating capital. Ramel and Michaelson (2020) apply a peer-based index of EUR corporate green and non-green bonds and report first evidence that green bonds outperform non-green bonds. A better performance of green bonds during periods of extreme market volatility and times of shock can be attributed to multiple reasons. First, investors may restructure their portfolios to reduce risk, thereby shifting towards financial instruments that offer detailed reporting and high transparency. These products can enhance trust and mitigate information asymmetry. Second, Environmental, Social, and Governance (ESG) considerations have gained increased attention during the pandemic. Investors have become more aware of the potential risks associated with environmental degradation and social inequality. Green bonds, which align with sustainable and responsible investing, might be perceived as a safer and more ethical investment option. In general, COVID-19 has acted as a catalyst for a broader shift in investor preferences towards sustainable investments.

To extend existing research, we do not use an index as utilized by Ramel and Michaelson (2020) but instead use matched pairs of bonds. We obtain data for 66 green bonds and compare the performance of these green bonds with their non-green peers. For every green bond, we therefore match synthetic conventional bonds using two bonds from the same issuer and the closest maturity, leading to 132 conventional bonds as a control group. The results between the two types of bonds indicate a highly significant negative premium of 1.6 bp for green bonds before the outbreak of COVID-19. Looking at the bonds during times of stress, the beginning of the pandemic, we see outperformance of the green bonds vs. conventional bonds, which concludes in a widening of the significant negative premium to 3.5 bp. The significant widening of the greenium also remains in a multivariate setting, controlling for other potential effects, or when changing our baseline regression model.

However, the mean greenium is perhaps a misleading indicator for financial managers who have to decide on the instrument to raise debt capital if the mean greenium is driven by a minority of issuers with outstanding financing conditions. To address this aspect, we focus on the credit ratings of corporate bond issuers and analyze whether the greenium for these issuers is different from the overall mean. We therefore extend the prior literature (e.g., Zerbib 2019; Hachenberg and Schiereck 2018; Löffler et al. 2021) and contribute to the understanding of the importance of credit ratings for firms.

In addition, we analyze whether the credit spread of a green bond is influenced by the sustainability of the country in which the green bond issuer is located. Stellner et al. (2015) analyze 872 bonds from 12 countries across the Economic and Monetary Union (EMU) and their respective zero-volatility spread (z-spread) as well as credit rating. They underline that the relationship between the corporate social performance of a company and the corresponding credit risk spread is moderated by the sustainability of the company's country of origin. The same holds for the credit rating, i.e., superior corporate social responsibility is rewarded in countries with above-average ESG performance. Investigating further the negative premium, we notice a moderating effect of the issuer's country environmental performance. The interaction term with the country's ESG performance indicates that the greenium was more pronounced for issuers from non-green countries prior to COVID-19. Examining the bond spreads since COVID-19, we do not find statistical difference between green and non-green countries.

This paper contributes to the growing literature on green bonds, showing how green bonds perform during financial distress conditional on the corporate credit rating and the issuer country's ESG performance. We find that they act as a convenient financing source which has important implications for practitioners. We also add to the current literature on sustainable investments (Albuquerque et al. 2020, Pastor et al. 2022), finding evidence that sustainable investments are more resilient during an exogenous market crash than conventional financial instruments.

Hypotheses development

Risk aversion regarding physical and transformational risk while the need for a sustainable transition grows may lead different groups of investors to re-think their portfolio structures in different ways. Times of stress may accelerate their behavior and may unveil divergences in the behavior of different investor groups to a larger extent. Collin-Dufresne et al. (2001) analyze credit spread changes of industrial bonds and conclude that credit spreads depend not only on traditional models but are also subject to local supply and demand shocks. As green and conventional bonds issued by the same company represent identical credit risk, supply and demand shocks triggered by COVID-19 may lead to changes in credit spreads between green bonds and their matched counterparts especially in the case of different groups of investors.

Immel et al. (2020) provide evidence that green bond issuers benefit from an average greenium of 8–13 bp. In addition, they find that green bond issuers are future-oriented companies whose governance structure is one of the drivers for lower issuance cost. In general, those companies may be better prepared to overcome a crisis and have the possibility to act in a more agile way, which should benefit their bond prices and cost of issuance. During risk-off periods, green bonds, which do not finance industries that have been hit particularly hard by the COVID-19 outbreak, like oil, gas, coal, and defense, should gain comparatively.

For conventional bonds, investors are usually unaware about the use of proceeds, leading to information asymmetries between firm and investors. The company issuing the bonds is not obliged to disclose the exact use of proceeds; use for “general corporate purposes” is a sufficient and frequent statement. Green bonds, on the contrary, give investors frequently the opportunity to follow the use of proceeds through regular reporting (ICMA 2021) and, in many cases, get certified by a second opinion. Harrison et al. (2020), e.g., find in a survey that around 85% of issuers make use of a second opinion for their first green bond. Thus, information asymmetries, which may concern investors, especially during times of high uncertainty, may be mitigated for green bonds and thus drive investors into these financial instruments. In addition, green bonds are frequently bought by buy-and-hold investors, such as insurance companies or pension funds (Harrison et al. 2020). Flammer (2021) analyzes investor characteristics following the issuance of green bonds. She explores that green bond issues attract institutional investors which are long-term oriented holders. She uses two different measures to account for long-term investment. First, she defines the holding horizon of investors. Therefore, she determines that an investor is long-term if the duration measure is above the median across all investors. Second, she uses the churn ratio and defines that an investor is long-term if the churn ratio is below the median across all investors. We expect investor groups invested in green bonds to act more resilient in times of stress than investor groups dominantly invested in conventional bonds. By analyzing green and conventional bonds, Bachelet et al. (2019) show that green bonds experience lower volatility than their matched peers during normal times. This leads us to our first hypothesis:

Hypothesis 1

During market stress, green bonds outperform conventional bonds, leading to a widening of the greenium compared to normal market periods.

Löffler et al. (2021), Bhurjee and Paliwal (2022), Hu et al. (2022) and Teti et al. (2022) concentrate their research not only on the existence of a greenium but also focus on potential determinants, indicating that the greenium might disperse between issues of different characteristics. Credit worthiness is one of the essential issuer characteristics in debt markets, as expressed in credit ratings. On the one hand, Hachenberg and Schiereck (2018), Zerbib (2019) as well as Janda and Zhang (2022) show that the greenium differs across rating classes. Löffler et al. (2021) find, on average, lower rated issuers for green compared to conventional bonds. On the other hand, it is regularly observed that credit spreads increase during volatile periods. This effect can be explained by risk averse investors who tend to move up the rating scale in volatile times and flee into higher rated issues. At the same time, lower rated issuers are confronted with a disproportionately high increase in debt financing cost. This increase might generate an additional incentive to spend effort in fulfilling the requirements of a green bond issue. This brings us to our second hypothesis:

Hypothesis 2

The greenium for issuers of green bonds is similar across credit rating classes.

Janda and Zhang (2022) recently documented that corporate ESG ratings have a significant impact on the green bond premium of corporate issuers. Monasterolo and Raberto (2018) show that sovereign green bonds promote greening the production systems and positively affect, among others, stable economic growth and emissions. Stellner et al. (2015) provide evidence for corporate issuers based in the EMU that these companies benefit from better country ESG ratings and lower z-spreads in their home countries. We combine the approaches of Janda and Zhang (2022) and Stellner et al. (2015) and test whether the green bond premium is more pronounced for firms headquartered in eco-friendly countries, leading to our last hypothesis:

Hypothesis 3

Issuers of green bonds based in an eco-friendly country benefit from a higher greenium than issuers based in less eco-friendly countries.

Data and methodology

The aim of this paper is, among others, the comparison of a potential greenium pre- and during-COVID-19. As matching pairs of bonds is the most accurate way of exploring a potential greenium, we follow the work of Hachenberg and Schiereck (2018) as well as Zerbib (2019) and compare green bonds with their (synthetic) conventional counterpart. All data for this analysis is obtained from Bloomberg as of November 17th, 2020.

First, we look at the universe of green-labeled bonds traded in the secondary market. To ensure that green and matched conventional bonds are as identical as possible, we exclude floating rate notes and callable bonds (despite bonds with make whole calls and calls at par three months before maturity). Next, we assure that our sample consists only of preferred senior, non-preferred senior (NPS), and senior unsecured bonds. We are interested in a potential greenium of corporate and financial bonds; thus, we exclude bonds issued by sovereigns, supranationals, and agencies (SSAs) as well as development and central banks. We focus on USD and Euro as issuing currencies and exclude all bonds whose issue size is less than 250 million or whose remaining maturity is shorter than 18 months to account for a liquid secondary market. The sample process leaves us with 341 green bonds. Table 1 summarizes the selection process.

Table 1 Sample selection process

In the next step, we have to find the comparable conventional bond to determine if a greenium exists. That bond must be issued by the same firm and must have the same characteristics as the green bond. We could follow Helwege et al. (2014) and use the conventional bond with the closest maturity, but according to Zerbib (2019), this may create a maturity bias. Thus, instead, we create a synthetic conventional bond by interpolating (or extrapolating) the two conventional bonds with the closest maturities to our green bond. The maturity of the conventional bonds should not differ more than two years from the green bond. This further reduces our data set to 104 green and 208 conventional bonds for the subsequent analyses.

Following Hachenberg and Schiereck (2018), we use the following formula to calculate the yield of the synthetic bonds:

$$\tilde{y}^{{{\text{NGB}}}} = y^{{{\text{NGB1}}}} + \frac{{y^{{{\text{NGB2}}}} - y^{{{\text{NGB1}}}} }}{{m^{{{\text{NGB2}}}} - m^{{{\text{NGB1}}}} }} \left( {m^{{{\text{GB}}}} - m^{{{\text{NGB1}}}} } \right)$$

(1)

where $\tilde{y}^{{{\text{NGB}}}}$ is the ask side of the yield of the synthetic conventional bond, $y^{{{\text{NGB}}1}}$ is the ask side of the yield of the conventional bond with the shorter maturity, $y^{{{\text{NGB}}2}}$ is the ask side of the yield of the conventional bond with the longer maturity, $m^{{{\text{NGB}}1}}$ is the maturity of the conventional bond with the shorter maturity, $m^{{{\text{NGB}}2}}$ the maturity of the conventional bond with the longer maturity, and $m^{{{\text{GB}}}}$ the remaining maturity of the green bond.

Differences in yield $\Delta \tilde{y}_{{{\text{it}}}}$ are calculated using the following formula:

$$\Delta \tilde{y}_{{{\text{it}}}} = y_{{{\text{it}}}}^{{{\text{GB}}}} - \tilde{y}_{{{\text{it}}}}^{{{\text{NGB}}}}$$

(2)

where $y_{{{\text{it}}}}^{{{\text{GB}}}}$ is the observed yield i of the green bond at trading day t and $\tilde{y}_{{{\text{it}}}}^{{{\text{NGB}}}}$ the observed yield i of the synthetic conventional bond at trading day t.

Thereafter, we distinguish between a regular market period and a market period highly influenced by an exogenous shock, COVID-19. We test if a potential greenium during a standard market period differs from a greenium during a market sell-off and a period of exceptionally high volatility. Fear of a worldwide pandemic hit equity markets on the 24th of February 2020, with the S&P 500 index falling by 3.35% on that day, the highest loss in two years. Credit derivatives markets mirrored the picture, with 5y iTraxx Main widening from around 50 bp before the pandemic outbreak to nearly 140 bp in mid-March and 5y iTraxx Crossover rising from close to 200 to 626 bp between January and March (Reuters 2020). The volatility around this time is further supported by academic evidence: Ramelli and Wagner (2020) show that the health crisis due to COVID-19 was amplified through financial channels, while Baker et al. (2020) document a drastic spike in economic uncertainty in March 2020 proxied by market volatility, newspaper articles and surveys.

We define the first-time horizon, the normal market period, from 3rd of June 2019 to 13th of February 2020. The second period, marked by a volatility last seen during the financial crisis of 2008/2009, an intense market sell-off, and a flight to quality, we define from 27th of February 2020 to 25th of March 2020. To assure liquidity of the sample, we do not include bonds (green and conventional) with a remaining maturity of fewer than 24 months. This further reduces our data to 66 green bonds and a total of 10,160 trading days for the first (p1) and 60 green bonds with 1,200 observations for the second period (p2). In order to test for significance of a potential greenium, we conduct the parametric t-test and non-parametric Wilcoxon rank-sum test for both time periods.²

After testing for significance of the potential greenium, we are interested if the greenium differs in period 1 (regular market period) from period 2 (exogenous shock through COVID-19). Therefore, we use a random-effects regression, with the greenium $\Delta \tilde{y}_{{{\text{it}}}}$ as the dependent variable and include a binary variable period 1 that takes the value of 1 if the observation is from period 1, and value 0 if the observation is out of period 2 as well as several control variables.³${\mathrm{Fin}}_{i}$ is a binary variable, which takes the value of 1 if the issuer of the green bond is a financial company and 0 otherwise, ${\mathrm{Maturity}}_{i}$ the remaining maturity of the green bond i in years, and ${\mathrm{Rating}}_{i}$ the rating of the green bond on a 22-step numerical scale (AAA = 22, AA + = 21,…, D = 1). Ratings of the green bonds are retrieved from Bloomberg. We decide to use the “Bloomberg Composite Rating” which is the mean of the ratings provided by the three main rating agencies, namely Moody's, Standard & Poor's, and Fitch. Log (Size)_i is the logarithmized issue size of the green bond, which leads us finally to our main model stated in formula (3):

$$\begin{aligned} \Delta \tilde{y}_{{{\text{it}}}} &= \beta_{0} + \beta_{1} {\text{p1}}_{i} + \beta_{2} {\text{Fin}}_{i} + \beta_{3} {\text{Maturity}}_{i} + \beta_{4} {\text{rating}}_{i} \\ &\quad+ \beta_{5} {\text{LOG}}\left( {{\text{SizeGB}}} \right)_{i} + \varepsilon_{{{\text{it}}}} \end{aligned}$$

(3)

In order to check for Hypothesis 3 and to control whether the firm's country of origin has an impact on a potential greenium, we introduce the variable green country. To measure the level of the green country, we apply the Environmental Performance Index (EPI) published by Yale University. EPI uses quantitative data to compare the environmental health and ecosystem vitality of 180 countries. We apply the following random-effects panel regression:

$$\begin{aligned} \Delta \tilde{y}_{it} & = \beta_{0} + \beta_{1} {\text{green country}}_{i} + \beta_{2} {\text{MSCI AAA}}_{i} \\ & \quad+ \beta_{3} {\text{MSCI AA}}_{i} + \beta_{4} {\text{MSCI A}}_{i} + \beta_{5} {\text{MSCI BBB}}_{i}\\ &\quad+ \beta_{6} {\text{Disclosure}}_{i} + {\text{Controls}} + \varepsilon_{{{\text{it}}}} \\ \end{aligned}$$

(4)

where $\Delta \tilde{y}_{it}$ is the greenium and green country is our primary variable of interest. ${\text{green country}}_{i}$ is a binary variable, which takes the value of 1 if the country of the issuer of the bond is based in one of the five highest scored countries by EPI, and is 0 otherwise. These countries were at that time Denmark, Luxembourg, Switzerland, Great Britain, and France. We additionally control for several other effects that potentially affect the greenium by including multiple control variables and our initial variables. First, we control for the ESG rating of the bond issuer provided by MSCI. ${\text{MSCI AAA}}_{i}$ is a binary variable, which takes the value of 1 if the issuer owns an MSCI ESG rating of AAA, 0 otherwise. ${\text{MSCI AA}}_{i}$ is a binary variable, which takes the value 1 if the issuer of the bond has an MSCI ESG rating of AA, 0 otherwise. Likewise, ${\text{MSCI A}}_{i}$ is a binary variable, which takes the value of 1 if the issuer's ESG rating by MSCI is A, 0 otherwise, and ${\text{MSCI BBB}}_{i}$ is a binary variable, which takes the value of 1 if the bond issuer owns an MSCI ESG rating of BBB, and 0 otherwise. We also include a control variable that measures the willingness of the issuer to provide ESG data. Therefore, we apply the Bloomberg disclosure score. The dummy variable ${\text{Disclosure}}_{i}$ takes the value of 1 if Bloomberg provides a disclosure score for the issuer (which is only the case if the issuer already supplied ESG data), and 0 otherwise. Last, we include for robustness our variables from the first regression. The variables used throughout the paper are outlined in Table 2, while the descriptive statistics are provided in Table 3 and split in the pre COVID-19 time (Panel A) and during COVID-19 (Panel B).

Table 2 Definitions of variables

Table 3 Summarized statistics

Results

First, we analyze green financial and corporate bonds and their respective synthetic conventional counterpart as we are interested in a potential greenium. The results of the univariate analyses are presented in Table 4. We find a significant greenium during both periods, 1.6 bp during the first period, before the outbreak of the pandemic, and 3.5 bp in the second period, during COVID-19. The difference between the two periods is nearly 2 bps and the greenium is considerably more pronounced during COVID-19. This shows that during market stress, green bond investors who are assumed to be long-term oriented investors act more resilient and confirms the findings of Harrison et al. (2020) as well as Bachelet et al. (2019) that show greater stability to secondary market performance in times of volatility. Both tests, the Wilcoxon rank-sum as well as the t-test, show significance, which is at the 5% level for the first-time frame and at the 1% level for the second time frame.

Table 4 Univariate results significance of greenium

Second, we examine if the greenium for green bonds traded during regular market times (period 1) differs from the greenium for green bonds traded during times of extreme volatility, the COVID-19 pandemic (period 2) using regression analyses. We test the greenium with a random-effects regression and introduce a binary variable for the normal market period. The results of our model are provided in Table 5 and show that the dummy variable for period 1 (which has the value one if the green bond was traded during the normal market period, pre COVID-19 and zero if it was traded during times of stress and the outbreak of COVID-19) is highly significant. Comparing markets pre COVID-19 and during COVID-19, the greenium widened as the coefficient is significant and positive, supporting the results of the univariate regression but controlling for several other effects.⁴ As the greenium shows a negative green bond premium, i.e., green bonds are traded tighter than conventional bonds, a widening premium during period 2 indicates flight into green bonds during times of high market stress and volatility. Green bonds are bought and act as a safe haven compared to conventional bonds issued by the same companies and owning the same bond characteristics. The control variable for the credit rating included testing Hypothesis 2 has a negative impact on the greenium. As we defined the best credit rating “AAA” as the highest number, the positive coefficient has a negative effect on the greenium, indicating that the better the rating, the smaller the greenium. This finding is of high practical importance as it states that the greenium varies among rating classes, and most green bond issuers generated a smaller greenium than the mean indicated. Note that while we do not report the results for reasons of brevity, we also find the effect for the two most common letters “A” and “BBB” using two dummy variables for the two ratings instead of the continuous variable.

Table 5 Regression results greenium pre COVID-19 and during COVID-19

In line with recent studies, the results provide evidence supporting Hypothesis 1. Our analyzed set of green bonds trades at a greenium compared to their synthetic counterparts (bonds with the same characteristics). During times of extreme market volatility triggered by the worldwide pandemic COVID-19, this greenium widens, i.e., green bonds outperform conventional, non-green bonds issued by the same entities.

Next, we analyze whether the eco-friendliness of the country the issuer of the bond is based in (measured through EPI) has a significant effect on the greenium. The results of the model are provided in Table 6. In model 1, we first introduce the variable ${\text{green country }}$ and other controlling variables. While we can confirm our prior results on the credit rating and the lower greenium in period 1, the coefficient for the green country variable lacks significance. In model 2, we interact the period dummy with the green country variable and the result indicates that the interaction term is now significant and positive, suggesting that the impact is particularly driven by both effects and the greenium effect is less pronounced for issuers from green countries prior to COVID-19. Model 3 and 4 splits the time intervals and the results support the prior models that the greenium is significantly more pronounced in the second period for firms in green countries. Thus, we have to partly reject our Hypothesis 3 as we do not find a continuous effect for the country, but only for the first period. Our control variables do not outline any consistent significances. Disclosure of data, i.e., transparency, is not significant at all. The second best ESG rating by MSCI, AA, is significant at the 10% level in the second period but not before. We also find that the dummy for the financial industry is significant in the first period, but lacks significance during the COVID-19 period.

Table 6 Regression results green country and control variables

Finally, even though we find that the random effects model is appropriate for our setting testing the greenium before and during COVID-19, we nevertheless ran as a robustness test the same model using a fixed effects model with country fixed effects. The results using a fixed effects model are provided in Table 7 and show that our results are not driven by the choice of the model. The coefficient of pre-Covid is (compared to the base category of during Covid-19) even more pronounced than in the regression using random fixed effects, suggesting that the results reported in the baseline regression are more cautious.⁵

Table 7 Regression results on greenium pre COVID-19 and during COVID-19

Conclusion

In this paper, we examine green bonds and their conventional counterparts during regular market periods, before the outbreak of the worldwide pandemic caused by the COVID-19. The yields of 10,160 observations show a negative premium of 1.6 bp during normal times, i.e., green bonds trade tighter than their conventional peers. This greenium is highly significant and consistent with the findings of Gianfrate and Peri (2019), Zerbib (2019) and Hachenberg and Schiereck (2018). In times of high market volatility and stress, during the outbreak of COVID-19, this greenium is persistent but widens to 3.5 bp during the COVID-19 period. Our regression analysis confirms the significant increase in the greenium, controlling for multiple additional factors. Green bonds benefit from a market sell-off, representing a convenient financing source. While the market suffers from an exogenous shock, the greenium is still significant. We conclude that green bonds outperform their non-green peers in extreme market situations, such as the COVID-19 pandemic and the financial turmoil related to lockdowns and slowing down of the global economy. However, as an avenue for future research the robustness of our findings needs future verification.

Including a binary variable that separates regular markets, before the outbreak of COVID-19 from markets of high stress and volatility, during COVID-19, it becomes evident that this variable is highly significant and the greenium value dependent on the respective market period. Albuquerque et al. (2020) and Pastor and Vorsatz (2020) show that sustainable financial instruments, stocks as well as mutual funds are more resilient to market stress than their conventional peers. Our results further support their findings.

However, the sole focus on a mean greenium is a misleading indicator for financial managers who have to decide on the instrument to raise debt capital. We show that the greenium also depends to some extent on the credit rating of the issuer as firms with better credit ratings show a smaller greenium effect. Note that this finding can only be interpreted as preliminary evidence as Agliardi and Agliardi (2021) show that green bonds affect the issuer's creditworthiness, depending on the correlation between green projects and the core business. This can induce additional effects for later issues.

Stellner et al. (2015) provide evidence that country sustainability affects spreads as well as credit ratings of corporate bonds. We add to this research by showing that the geographic effect on the greenium persists only during the time without financial distress using an interaction variable. Throughout periods of high stress, the eco-friendliness of the issuer's country of origin did not influence the greenium significantly. Note that our results are mixed as by splitting the two time intervals, the results for the green country lack significance in both periods. We used different investigation periods and also our time period of normal times is different to Stellner et al. (2015), which might explain the differences in the results.