(In)-Credibly Green: Which Bonds Trade at a Green Bond Premium? · 2019-07-15 · (In)-Credibly...
Transcript of (In)-Credibly Green: Which Bonds Trade at a Green Bond Premium? · 2019-07-15 · (In)-Credibly...
(In)-Credibly Green: Which Bonds Trade at a
Green Bond Premium?∗
Julia Kapraun† & Christopher Scheins‡
This version: May 2, 2019
Abstract This paper provides an in-depth analysis of the green bond premium, so-called
Greenium, using both, primary and secondary market data. We consider a large sample of
over 1,500 green bonds issued worldwide and estimate the di�erences in yields of green and
comparable conventional bonds. Our primary market results reveal a signi�cantly negative
premium of 20�30 bps for green bonds, implying that at issuance, green bonds are trading
at lower yields than their conventional counterparts. This premium, however, varies across
currencies and issuer types and over time. In particular, credibility plays an important role
as bonds issued by more credible entities or bonds denominated in major currencies are issued
at lower yields. Using secondary markets data we �nd green bonds trading at higher yields
except for those issued by governments and supranational institutions. Further, bonds listed on
exchanges with a dedicated green bond segment trade at up to 13 bps lower yields, pointing out
the importance of transparency and clear standards for the growth of the green bonds market.
Finally, we reveal the relevance of the overall sustainability reputation of the issuer for the
acceptance of green bonds by investors. Here, we �nd the di�erence in yields between green
and conventional bonds to be higher (i.e., lower prices for green bonds) for the issuers coming
from both, the top and the bottom 30% based on the overall ESG rating, which might be due
to investors fearing green�labeling or green-washing e�ects.
Keywords: Green bonds, Sustainable investing, Green premium, Green exchanges
JEL Classi�cation: C33, G12, G14, G20, Q56
∗We thank Climate Bond Initiative for providing access to their data set and supporting the researchproject. We are also grateful for comments from Christian Schlag, Steven Ongena, Hannes Wagnerand Andreas Hoepner, and participants of the research seminars at Goethe University Frankfurt, Zurichuniversity, the EU conference �Promoting Sustainable Finance� in Brussels and Climate Bond Initiativeannual conference in London. Any errors, misrepresentations, and omissions are our own.†Corresponding author. Goethe-Universität Frankfurt, Theodor-W.-Adorno-Platz 3, 60323 Frankfurt,
email: [email protected]. Tel.: +49(0)69798 33728‡Goethe-Universität Frankfurt, Theodor-W.-Adorno-Platz 3, 60323 Frankfurt, email:
1 Introduction
The conference on climate change held in Paris in 2015 paved the path for environmen-
tally sustainable growth and set a road map for countries to follow in order to keep the
average temperature from rising by more than 2 degrees Celsius by 2030. Three years
later, the most recent Emissions Gap Report published by UN Environment Programme1
revealed this target to appear illusory. Not only that most countries stayed far below their
Nationally Determined Contributions, but the total annual greenhouse gases emissions
reached a record high in 2017. To achieve the temperature goals of the Paris Agreement
countries will need to triple their e�orts and increase e�ectiveness of domestic policy.
Within this framework, green bonds, whose use of proceeds is earmarked to �nance or re-
�nance environmentally sustainable projects, are seen as one of the key catalysts bearing
the potential to drive the shift to a low-carbon global economy. Glomsrød and Wei (2018)
estimated that the diversion of capital from fossil industries to the green �nance sectors,
in particular through green bonds, would increase the world GDP and drastically reduce
the GHG emissions. Further evidence on the potential of green bonds in guaranteeing
economic sustainable growth is provided by Flaherty et al. (2017).
Our study is motivated by the ongoing debate about the credibility of green bonds and
the willingness of investors to pay a higher price for a green bond compared to the
conventional one. Our main research questions are (a) is there a di�erence in yields
between green and conventional bonds; (b) is there a variation across investors (primary vs
secondary markets), markets (currencies), issuers (corporates vs governments) and time;
and (c) are there any green bonds, which appear to be more reliable (�greencredible�) and
thus, trade at higher prices.
Over the past few years, the global market for green bonds increased signi�cantly and
reached 2017 volumes of about $155 bn, issued by governments, supranationals, and cor-
porates from di�erent sectors in over 30 di�erent currencies. With the rapidly growing
number of green bond issuers across the globe, the need for more transparency, regulation
1https://www.unenvironment.org/resources/emissions-gap-report-2018
1
and clear standards increases. Issuing green bonds is still a non-standardized process, and
each issuer and country can decide on the criteria making its bonds green. Though the
green bond principles have become the leading framework for issuance of green bonds2,
and EU Commission is working on a Proposal for a EU Green Bond Standard, these
guidelines are still voluntary, and there is no harmonized framework for eligible green
projects or yearly reporting. Many investors fear labeling e�ects or even green wash-
ing and question the �greencredibility� of issuers, which have a reputation of the major
polluters (e.g., from sectors such as transportation or nuclear power producers) or coun-
tries with the lowest environmental liability3. Though some studies �nd green bonds
value-enhancing and attracting new investor clientele (e.g., Flammer (2018)), based on
interviews with practitioners, Reed et al. (2019) claims green bonds to have not realized
a price premium due to investors not trusting the green label.
To investigate the existence of the green bond premium, so-called Greenium, we rely
on the analysis of both, primary and secondary market data across the full spectrum of
global green bonds. To the best of our knowledge, we are the �rst to collect and analyze a
large amount of the most recent and complete data on green bonds from di�erent sources.
Our initial data set contains 2,114 green bonds with a total issuance value of $487 bn with
the majority of these bonds issued within the last 2 years. To assess the pricing e�ect
of the green label, we, �rst, estimate the di�erence in the Yield at Issuance for green
and conventional bonds by controlling for di�erent bond characteristics and time e�ects.
Thereby we enlarge the available data set by computing the issuance yield from other
bond characteristics. This allows us to consider 1,520 green and 202,394 conventional
bonds in a �xed e�ects regression analysis, where we regress the Yield at Issuance on a
green dummy variable and di�erent controls.
In a next step, we consider bond prices and yields on the secondary market. While
the results on the primary markets can be driven by the strong institutional demand,
2See https://www.icmagroup.org/green-social-and-sustainability-bonds/green-bond-principles-gbp/ forthe most recent guidelines
3https://www.theguardian.com/world/2018/apr/17/poland-violated-eu-laws-by-logging-in-biaowieza-forest-says-ecj
2
sustainable investments have not reached same dimensions among private investors4. In
particular, these investors might question the greencredibility of di�erent issuers or not be
willing to pay a higher price for a green bond of the same issuer. To disentangle the green-
ness e�ect from the issuer speci�c e�ects, we match every green bond with corresponding
conventional bonds of the same issuer, same currency, same coupon type and same se-
niority. We then estimate the di�erence in yields for every green-conventional bond pair.
In contrast to other studies, we do not rely on option-based data or construct synthetic
conventional bond data, which both signi�cantly reduce the data sample. Our matching
method allows us to compare 4,609 bond pairs. Finally, we test for the credibility channel
to reveal possible reputational e�ects on the green bonds market.
Our study provide an in-depth analysis of several key issues with regards to the pricing of
green bonds. The results from primary market reveal a signi�cantly negative premium of
20− 30 bps for green bonds, implying that at issuance, green bonds are trading at lower
yields than their conventional counterparts. The e�ect is much stronger for green bonds
issued in USD than in EUR (41bps vs 17bps) and not signi�cant for green bonds issued in
CNY. We further observe varying premia across di�erent issuer types. In particular, we
�nd very high negative premia for green bonds issued by more reliable issuance entities
such as governments or supranationals, while the premia for corporate green bonds are
much smaller or even not signi�cant anymore after controlling for di�erent issue size
e�ects. We further �nd a lot of variation in premia over time. While there are months
in a raw with signi�cantly lower yields for green bonds during the hype period after
the Paris Agreement 2016 to 2017, in the last 2 years, there are only few months with
a non-zero green bond premium, and these usually are due to a single large issuance
by a government or supranational entity (e.g., government bonds issued by France in
January 2017 or Belgium in February 2018). This result indicates that though, through
the high institutional demand on the primary markets, green bonds trade at a higher
price than conventional ones on average, reputation aspects should be taken into account
when assessing the pricing implications for green bonds. These aspects are particularly
4So for instance, Union Investment reports 38bn euro in institutional portfolios, while there are only 3bnin private portfolios (see Boersenzeitung on February 15, 2019).
3
relevant in our secondary market analysis, where we even observe a positive premium for
green bonds. Here only green bonds issued by governments or supranational institutions
are traded at a small but signi�cant premium of -2 bps, while corporate bonds have on
average 43 bps higher yields than conventional bonds. This �nding clearly underlines the
importance of the issuer's credibility for the greenness reputation of the bond. Further,
we �nd the di�erence in yields between green and conventional bonds to be higher for
more liquid bonds with larger issue sizes. Most interestingly, we observe a signi�cant
e�ect of the listing of the bond, as green bonds listed on an exchange with a dedicated
green market segment (i.e., Luxembourg or London Stock Exchange) trade at up to 13
bps lower yields, on average. This indicates that green bonds on dedicated markets
appear to be more credible through additional visibility and transparency, and reveals
the importance of exchanges for the market growth. Finally, we reveal the relevance of
the overall sustainability reputation of the corporate issuer for the acceptance of green
bonds by investors. Here, we �nd the di�erence in yields between green and conventional
bonds to be higher (i.e., lower prices for green bonds) for the issuers coming from the
top and bottom 30% based on the overall ESG rating. A rationale explanation for this
�nding might be that on the one hand, investors fear green�washing e�ects for bonds
coming from companies with a rather bad ESG reputation and green labeling e�ects for
bonds issued by companies with a very good sustainability reputation.
Our paper contributes and adds to the growing literature on the pricing impact of in-
vestor's preferences for sustainable assets. Many recent studies provide evidence that
investors generally value sustainability (Hartzmark and Sussman (2019)) and are willing
to pay for non-pecuniary characteristics of investments (Barber et al. (2018)). The results
on the pricing implications for green bonds are, however, mixed. Some few studies report
green bonds trading at a lower yield than conventional bonds. For instance, Ehlers and
Packer (2017) analyze the credit spread at issuance of 21 green bonds; Nanayakkara and
Colombage (2018) or Preclaw and Bakshi (2015) apply option-adjusted spread data to
measure the credit spreads of green and conventional bonds; Karpf and Mandel (2018)
use Oaxaca-Blinder regression for municipal green bonds and �nd the premium to be
4
negative from 2015 on; or most recently, Zerbib (2019) reports a small negative green
bond premium of -2 bps for a global set of green bonds on the secondary market and
Baker et al. (2018) �nd higher prices for green municipal bonds at issuance. Other studies
(e.g., Karpf and Mandel (2017), Hachenberg and Schiereck (2018) or reports from CBI5,
or other practitioner's reports) document a contrary �nding on the sign of the premium
or even no signi�cant di�erence in yields. However, most of these studies rely only on
a very small set of bonds (less than 150, which accounts for less than 10% of the whole
sample) issued before the hype issuance years 2017 and 2018. Further, some do not take
time e�ects or other relevant bond characteristics such as currency or issuer characteris-
tics into account. Finally, none of the studies so far provided a full picture of the pricing
implication on both, primary and secondary markets or explored the greencredibility
channel.
2 Green bonds market
For this study we collect a large amount of data on green bonds from three di�erent
sources6. The initial data set contains 2,114 green bonds7, with a total issuance value of
$487 bn. A brief overview over some characteristics of the green bonds in our data set
can be found in Table 1. Around 75% of all green bonds in our sample have an issuance
volume below $350m. The average green bond is issued at a premium, with an average
maturity of 8.75 years, coupon of 3.40% and an issuance yield of 3.27%.
In 2007 the European Investment Bank pioneered the Green Bonds market by issuing
the �rst Climate Awareness Bond. 10 years later, more than 2,000 Green Bonds have
been issued in 30 di�erent currencies worldwide. The majority of these bonds followed
the Paris Agreement of 2015 with over 300 bonds issued yearly with increasing issuance
volumes. Figure 1 reports the yearly issuance volumes split per issuer type.
5See https://www.climatebonds.net/resources/reports/20186A list provided by CBI; Bonds from Bloomberg whose �Use of Proceeds� contains Green; Bonds fromReuters that are classi�ed as green
7We excluded more than thousand mortgage bonds issued by Fannie Mae with the average size of $23m
5
25% 50% Mean 75% 95% N
Coupon(%) 1.38 3.25 3.40 5.00 8.00 2,069Experienced 0.00 1.00 0.64 1.00 1.00 2,114GreenEX 0.00 0.00 0.21 0.00 1.00 2,114Issue Price 99.86 100.00 100.40 100.00 111.03 1,828Issue Yield 1.49 3.01 3.27 4.69 7.80 1,787Maturity(Years) 4.00 5.01 8.75 10.01 29.82 2,100Volume($Million) 11.02 65.55 234.22 302.07 890.04 2,111
Table 1: Descriptives of our dataset on green bonds.Coupon (%) is the size of the annual coupon of the bonds; Experienced is a dummy variable that isone if the bond is issued not on the same day as the �rst green bond of this issuer; greenEX is a dummyvariable that is one if the bond is listed on a green exchange; Issue Price is the price of the bond atissuance; Issue Yield is the yield to maturity of the bond at issuance, which is computed if no data wasprovided by Reuters; Maturity (Years) is the di�erence between the Issue Date and Maturity Date inYears; Volume ($Million) is the issue volume of the bond converted in million USD.
2010 2012 2014 2016 2018
Year
0
20
40
60
80
100
120
140
bnUSD
CORP
SOVR
SUPR
Figure 1: Volume of green bonds by issuer type
While the �rst green bonds were issued by supranationals8, by the end of 2018, corpora-
tions from di�erent sectors (e.g., energy, �nancials) and government entities are among
the largest issuers. Table 2 displays the top 10 Issuers with respect to the number of
bonds issued and issuance volumes. Nearly half of the corporate green bonds come from
8European Investment Bank (EIB) and World Bank
6
Issuer bnUSD # Issued Sector
European Investment Bank 40.19 49 Banking ServicesWindMW GmbH 29.37 80 Electric Utilities & IPPsÉlectricité de France S.A. 26.19 18 Multiline UtilitiesIBRD 17.00 158 Banking ServicesIndustrial Bank Co Ltd 16.93 7 Banking ServicesKfW 14.74 18 Banking ServicesMexico City Airport Trust 12.00 8 Collective InvestmentsInternational Finance Corp 8.46 77 Investment Banking & Investment ServicesNRW Bank 7.86 10 Banking ServicesShanghai Pudong Development Bank Co Ltd 7.59 3 Banking Services
Table 2: Top 10 Green Bond Issuer.Issuer is the most frequently used name for each Parent Identi�er, bnUSD is the USD equivalentissue volume, Number Issued is the number of Green Bonds that were issued by the Parent Identi�er,Sector is the most frequently named sector for each Parent Identi�er.
0-19 20-39 40-59 60-79 80-100
ESG Score
0
10
20
30
40
50
60
70
Frequency
Figure 2: ESG ratings for green bonds issuers
Banks, followed by Corporate Financial Services and Electric Utilities sector. Interest-
ingly, while most of these issuers have a solid reputation of sustainability in general,
there are several green bonds coming from companies with very low sustainability rank-
ings. Figure 2 depicts a histogram for ESG ratings based on available data for 126 issuers
from Thomson Reuters database. Thereby, the average company has an average score of
63.68 (score between 0 and 100), 25% of the issuers have a score below 53 and 10% have
a score above 83.
7
Alternative Energy
8%
Clean Transport
15%
Energy E�ciency
18%
Other23%
Eligible Green Bond Projects
35%
Green Amount Issued (bnUSD)
Alternative Energy
13%
Clean Transport
13%
Energy E�ciency
11%
Other
39%
Eligible Green Bond Projects
25%
Green Number Issued
Figure 3: Use of Proceeds of green bonds
Most of the green bonds from these issuers �nance renewable energy projects, energy
e�cient buildings or used to �nance clean transport and sustainable water management
projects, as can be seen from Figure 3.
Green Bonds are issued in every major currency across the globe, though there is a
strong focus on the Americas, Asia, and Europe. Figure 4 presents the number of bonds
and issuance for di�erent currency markets. Most and largest bonds are issued in USD
($157bn), Euro ($146bn), Chinese Yuan ($75bn) or the Swedish Krona ($23bn). Though
there are more than twice as many green bonds issued in USD compared to those issued
in EUR, these have lower issuance volumes, on average. With 16 per cent of the world's
green bonds China is emerging as a key player in the green bond market, though the
Chinese guidelines on the green bond issuance deviate from the international standard.
In particular, only 50% of proceeds are required to be invested in sustainable projects,
while it is the case for over 90% in Europe.9
In our sample we have in total 559 di�erent issuers. Figure 5 displays a histogram of the
9See https://www.bourse.lu/sustainability_standards_and_labels for di�erent green bond stan-dards.
8
SEK5%
Other
13%
CNY
16%
EUR 33%
USD
34%
Green Amount Issued (bnUSD)
SEK11%
Other
30%
CNY9%
EUR
15%
USD
35%
Green Number Issued
Figure 4: Green Bonds issued by currency
number of green bonds per issuer. More than 50% of the issuers in our sample have issued
one single Green Bond so far, while 5% issuers account for 50% of all issued bonds. 36%
of all issuers have only Green Bonds, which will be particularly relevant for our analysis
of the secondary bond market, where we compare Green and conventional bonds of the
same issuer.
In the last few years many exchanges have launched dedicated segments exclusively for
green bonds, which improved the liquidity and transparency of the green bonds market
and provided access to di�erent types of investors. More importantly, Green exchanges
increase credibility of the Green label as bonds listed in dedicated segments are usually
required to meet certain standards with respect to reporting, external reviews etc. For
instance, Luxembourg Green exchange, one of the �rst dedicated green bond platforms,
require issuers to align with Green Bond Principles, the CBI Climate Bonds Standard
eligibility taxonomy or another related frameworks10. Table 3 provides an overview of
green bonds from our sample listed on such Green exchanges.11
10See https://www.bourse.lu/sustainability_standards_and_labels for an overview of variousgreen bond standards.
11The list of Exchanges with a green bond segment is, however, not complete, as we only have information
9
1 2-4 5-10 >10
Number of Issues
0
50
100
150
200
250
300Frequency
Figure 5: Number of Green Bonds per Issuer
Exchange Number bnUSD
Luxembourg Stock Exchange 113 56.22Frankfurt Stock Exchange 89 53.44Milan Stock Exchange 39 34.40London Stock Exchange 59 11.17Stockholm Stock Exchange 61 4.24Oslo Stock Exchange 12 2.13Taipei Exchange 15 1.33
Table 3: Number and Issuance Volume of Green Bonds in our sample that are traded onthe Green Exchanges
3 Data and Methodology
3.1 Data
One of the most important steps of our analysis is the collection of all relevant data
on Green and conventional bonds and their issuers. To this end, we �rst obtain a list
on the bonds in our sample per end of 2018 with existing information on the listing.
10
of ISINs of Green bonds, by taking the union of a list provided by the Climate Bond
Initiative (CBI), bonds that are classi�ed as green in Reuters and bonds with �green�
use of proceeds �ag on Bloomberg. Over the last years, Fannie Mae, a US government-
owned entity that provides a secondary market for home mortgages, issued a lot of green
bonds, however we could not obtain any trading characteristics for the most of them. In
addition, these bonds are usually relatively small and, therefore, we excluded all of them
for our analysis. Our �nal set consists of 2,114 Green bonds.
Next, we collect ISINs of conventional bonds with di�erent coupon types, issued after
2009 and which started with all combinations that are present in the ISO 31661 alpha-2
list12. Though we focus on Fixed Coupon Plain Vanilla Bonds (FXPV) in our analysis,
we also included other bond types for issuers, which also issued at least one green bond.
This gives us a universe of 408,997 Green and conventional bonds in total.
Finally, we downloaded main characteristics of these bonds, i.e. the issue date, maturity
date, coupon, yield at issuance, amount issued, currency, sector, ratings from di�erent
agencies and others using Reuters and Bloomberg as our primary data sources. Summary
statistics for the bonds used in our primary and secondary analyses can be found in Table 5
and Table 6 in the following sections.
3.2 Primary market
To reveal possible di�erences in expected returns on Green and conventional bonds, in
this section we focus on Yields at Issue following Baker et al. (2018). Our basic panel
regression model is
Yi,t,b =7∑
k=1
αk + β ·Greeni,t,b +5∑
k=1
ck + εi,t,b (1)
12The �rst two letters of any ISIN can be used to identify most countries of origin with the help of theISO 31661 alpha-2 list, i.e. US is the two letter abbreviation for the United States and DE is the twoletter abbreviation for Germany. We extended the list with XS, as this abbreviation is frequently usedfor international issues
11
50% 75% 95% Mean Max
Conventional 1 2 5 1.98 157Green Bonds 1 1 2 1.32 33
Table 4: Number of bonds issued on the same day by the same issuer
where Yi,t,b is the yield at issuance of a bond, αk are Issuer, YearMonth, Currency, Rating,
Seniority, Maturity and Issue size �xed e�ects, and Greeni,t,b is our main variable, which
is 1 if the bond is labeled as green, and ck are di�erent controls. We include an indicator
variable Experienced which is one for all green bonds issued not on the same day as
the �rst green bond of the same issuer. Table 4 displays the summary statistics for the
number of bonds (Green and conventional) issued by companies on the same day. While
most of the companies issue only one green bond at the same time (mean of 1.32), there
are issuers which issue the whole range of green bonds (e.g., Quantum solar Park issued
33 bonds in October 2017). We further include 4 variables describing the most common
use of proceeds, i.e., Eligible Green Bond Projects, Clean Transport, Renewable
Energy and Energy E�ciency.
In this analysis we consider only bonds with available data on either Issue Yield or
Issue Price. We focus on Fixed Coupon Plain Vanilla Bonds, which allows us the exact
calculation of the Issue Yield using maturity, coupon rate and frequency and issue price
information. We end up with 1,520 Green and 202,394 conventional bonds in total. In
some speci�cations our sample is further reduced due to unavailable data on e.g., rating or
seniority. Table 5 provides descriptive statistics for the green (Panel A) and conventional
(Panel B) bonds used for this analysis.
The characteristics for green bonds are very much similar to our full sample in Table 1.
An average green bond in our sample has a maturity of 9 years, issuance volume of
$288m and is issued at a premium. 19% of all bonds are listed at Green exchanges. The
corresponding conventional bonds in Panel B have similar maturities and slightly higher
coupons and yields. Due to several extreme outliers the mean Issue Price and Issuance
volumes are signi�cantly higher than those for the green bonds, though when we compare
the 50 and 75 percentiles, the relationship changes.
12
Panel A: Descriptives of the Green Bonds
25% 50% Mean 75% 95% N
Coupon(%) 1.60 3.40 3.50 5.00 7.76 1,520Experienced 0.00 1.00 0.66 1.00 1.00 1,520GreenEX 0.00 0.00 0.19 0.00 1.00 1,520Issue Price 99.79 100.00 100.67 100.00 112.86 1,513Issue Yield 1.60 3.16 3.38 4.70 7.75 1,520Maturity(Years) 4.00 5.25 9.04 10.01 29.90 1,520Volume($Million) 15.07 99.92 288.13 426.49 1,000.00 1,518
Panel B: Descriptives of the Conventional Bonds
25% 50% Mean 75% 95% N
Coupon(%) 1.88 3.40 3.68 5.00 8.19 202,394Issue Price 100.00 100.00 140.59 100.00 115.21 198,282Issue Yield 1.69 3.00 3.46 4.50 8.21 202,394Maturity(Years) 3.01 5.56 7.63 10.01 20.01 202,354Volume($Million) 12.74 49.76 472.64 200.00 1,150.00 201,946
Table 5: Descriptives of Bonds used in our primary regressions.Coupon (%) is the size of the annual coupon of the bonds; Experienced is a dummy variable thatis one if the bond is issued not on the same day as the �rst green bond of this issuer; greenEX is adummy variable that is one if the bond is listed on a green exchange; Issue Price is the price of thebond at issuance; Issue Yield is the yield to maturity of the bond at issuance; Maturity (Years)is the di�erence between the Issue Date and Maturity Date in Years; Volume ($Million) is the issuevolume of the bond converted in Million USD.
13
In a sub-sample analysis we consider bonds issued in three major currencies to reveal
possible di�erences across markets and obtain 261 green bonds in EUR, 392 green bonds
in USD and 180 green bonds in CNY with available data on Issue Yield. We further
include a regression analysis for di�erent issuer types such as corporates, government
and supranational entities. Out of 1,520 green plain vanilla bonds 784 are issued by
corporates, 322 by supranationals such as e.g., European Investment Bank, and 414 by
government entities.
3.3 Secondary market
In our secondary market analysis, we study the determinants of the green bond premium,
here measured as the di�erence in the Mid quotes between Green and conventional bonds
yields, in more detail. One of the challenges in this analysis is to properly de�ne the
relevant data sample of Green and conventional bonds. To this end, for every green
bond we look for conventional bonds, issued by the same Issuer, with the same Rating,
Seniority, Currency and Bond Type. We focus on pairs of conventional and green bonds,
instead of triplets, as the Maturity di�erence between the green bond and the conventional
is constant and can be controlled for, by either using the Maturity di�erence or using pair
�xed e�ects. Further, we only take the comparable bonds that are closest with respect
to squared Maturity di�erences to the green bond to form a pair. Finally, we restrict
our list of comparables to contain at most 10 conventional bonds for each green bond.
This is mostly driven by data availability issues and the e�ort to have a similar weighting
of small issuers and issuers that have issued several thousand bonds in our sample13.
Previous approaches (e.g., used in Zerbib (2019) or Helwege et al. (2014)) use option-
based data or compare green bonds to synthetic conventional bonds. To create these
synthetic bonds, at least two comparable bonds are required. However, these methods
signi�cantly reduce the data sample and conclusions derived from less than 10% of the
sample might be questionable. In contrast, our methodology allows us to compare up to
13The biggest issuer in our sample has issued 16k bonds, with less than 1% of them classi�ed as Green,while around 300 issuers in our sample have issued only one green bond.
14
Panel A: Descriptives of the Green Bonds.
25% 50% Mean 75% 95% N
Coupon(%) 1.00 2.12 2.51 3.51 6.33 647Experienced 0.00 1.00 0.71 1.00 1.00 647GreenEX 0.00 0.00 0.32 1.00 1.00 647Issue Price 99.55 99.87 99.50 100.00 100.00 639Issue Yield 1.00 2.16 2.55 3.55 6.20 639Maturity(Years) 5.00 5.09 7.39 10.01 15.01 646Volume($Million) 112.44 407.30 441.21 511.82 1,250.00 647
Panel B: Descriptives of all conventional Bonds.
25% 50% Mean 75% 95% N
Coupon(%) 1.09 2.31 2.61 3.62 6.25 2,810Issue Price 99.51 99.89 99.33 100.00 100.16 2,707Issue Yield 1.34 2.52 2.78 3.80 6.25 2,702Maturity(Years) 5.00 7.01 8.87 10.01 27.69 2,812Volume($Million) 62.31 371.42 1,058.64 955.15 3,855.24 2,814
Table 6: Descriptives of Bonds that are used in our secondary market regression analysis.Coupon (%) is the size of the annual coupon of the bonds; Experienced is a dummy variable thatis one if the bond is issued not on the same day as the �rst green bond of this issuer; greenEX is adummy variable that is one if the bond is listed on a green exchange; Issue Price is the price of thebond at issuance; Issue Yield is the yield to maturity of the bond at issuance; Maturity (Years)is the di�erence between the Issue Date and Maturity Date in Years; Volume ($Million) is the issuevolume of the bond converted in Million USD. Note, that the conventional bonds can be used multipletimes and thus N does not match Table 10.
4,609 Green-conventional bond pairs and to control for di�erent �xed e�ects.
Thus the number of pairs we analyze is higher than in many other studies for several
reasons: we only need one comparable bond which allows issuers that only have few
bonds to appear in our sample; we only need bid and ask yield data for one comparable
bond. The latter is a rather important point, as we were able to obtain bid and ask yields
for only around one quarter of our pairs, through Reuters. For our analysis, we use the
mid quotes of both bonds in each pair and calculate the spread between the Green and
conventional bond, henceforth called GMC. Further variables that we calculate are the
bid-ask spread for green and conventional bonds, as well as the di�erences between the
bonds with respect to Maturity, Coupon, and Issue Size. Detailed statistics about the
15
available data used in our secondary sample can be found in Table 6.
To pursue the credibility channel we include an indicator variable greenEX which is 1
if the bond is Green and listed on an exchange with a dedicated green bonds market
segment. To take into account the ESG rating of an issuer, we create dummy variables
for di�erent ESG rating categories. To this end, we collect ESG ratings from Reuters (the
ESGScore) and from Bloomberg (the Sustainalytics rank, the Robecosam total stability
rank and the ESG disclosure score) for in total 122 issuers in our sample. We then
calculate the average of the four scores (or those that are available) for each issuer and
create a high (ESGHigh), a medium (ESGMedium) and low (ESGLow) dummy if
the average is in the top 30%, middle 40% or lower 30% of all issuers, respectively.
We obtain in total 777 green bonds with at least one comparable conventional bond issued
by the same company. 4,609 bond pairs result from each bond being compared to up to
10 out of 3,504 conventional bonds. Interestingly, green bonds issued by companies which
also issue conventional bonds, appear to have shorter maturities, much larger issuance
volumes and are issued at a discount in contrast to the average green bond from our
primary analysis in Table 5.
4 Results
4.1 Results primary market
Table 7 presents our main results for the total sample of 1,520 green bonds. After ac-
counting for di�erent Fixed E�ects (Models 2�4) all speci�cations support the prediction
that Green bonds are issued at a lower yield of 20 to 34 basis points than comparable
conventional bonds.
In Model (5) we consider pricing e�ects on bonds being issued after the very �rst Green
bond of the same issuer. Though insigni�cant, the negative coe�cient indicates that
bonds issued by experienced companies seem to appear more credible and trade at lower
yields at issuance. In Model (6) we control for di�erent types of Green projects and �nd
16
IssueYield
(1) (2) (3) (4) (5) (6)
Green −0.015 −0.198∗∗∗ −0.201∗∗∗ −0.338∗∗∗ −0.266∗∗∗ −0.457∗∗∗(0.060) (0.058) (0.063) (0.078) (0.067) (0.125)
Experienced −0.098(0.113)
Energy E�ciency −0.076(0.165)
Alternative Energy −0.035(0.179)
Eligible Green Bond Projects 0.158(0.137)
Clean Transport 0.194(0.205)
Issuer FE No Yes Yes Yes Yes YesYearMonth FE Yes Yes Yes Yes Yes YesCurrency FE No Yes Yes Yes Yes YesRating FE No No No Yes Yes NoSeniority FE No No Yes Yes Yes YesMaturity FE No No Yes Yes Yes YesIssue Size FE No No Yes Yes Yes Yes
Green Bonds 1,520 1,520 1,328 787 787 1,170Observations 203,914 203,914 165,631 82,357 82,357 53,032R2 0.017 0.771 0.789 0.672 0.672 0.753Adjusted R2 0.016 0.758 0.774 0.654 0.654 0.731
Table 7: Primary Market AnalysisThe table shows results of the Fixed E�ects regressions of bonds Yield at Issuance in basis points forGreen and conventional bonds. The Issue Yield of Green and conventional �xed coupon plain vanillabonds is regressed against a Green indicator, which is 1 if the Bond is a green bond and 0 otherwise.Experienced is a dummy variable that is one if the bond is issued not on the same day as the �rst Greenbond of this issuer; Energy E�ciency, Renewable Energy, Clean Transport and Eligible GreenBond Project are dummy variables indicating the use of proceeds for the corresponding bond. Weinclude Issuer �xed e�ects, YearMonth �xed e�ects, Currency �xed e�ects, Rating, Seniority, Maturityand Issue size �xed e�ects, to take into account substantial di�erences between issuers, the yield curve,di�erent interest rate environments in di�erent countries and the in�uence of ratings on the yield atissuance. The Maturity �xed e�ects are three buckets for short-term (less than 5 years), medium (between5 and 10 years) and long term (more than 10 years) time to Maturity. The Issue Size �xed e�ects arethe deciles of the issue (in USD) in comparison to all other issues that occurred before or in the samemonth as the Issue.
these also to be insigni�cant, though the coe�cient on the Green variable doubles in
comparison to Model (3). We further observe that categories Energy E�ciency and
Renewable Energy seem to be more accepted as green projects thanClean Transport
or a rather general category of Eligible Green Bond Project.
17
To shed more light on the di�erent currency and issuer type e�ects Table 8 presents
results of Model (3) of Table 7 for di�erent sub-samples. First, in Models (1) to (3) we
observe a strong variation in the premia across currencies. While the coe�cient on the
Green dummy for EUR denominated bonds is similar to the full sample (17bps), the
result for USD denominated bonds is more than double as high (41bps). Interestingly,
Green Bonds issued in Chinese currency CNY appear to be less attractive and trade at
similar yields as their conventional counterparts. This might be an indication that Green
bonds, which follow international (and not local Chinese) Green Bond Principles are more
recognized among investors.
IssueYield
(1) (2) (3) (4) (5)
Green −0.171∗∗∗ −0.410∗∗∗ 0.053 0.053 −0.310∗∗∗(0.052) (0.077) (0.050) (0.068) (0.101)
Subsample EUR USD CNY CORP SOVR+SUPRIssuer FE Yes Yes Yes Yes YesYearMonth FE Yes Yes Yes Yes YesCurrency FE No No No Yes YesSeniority FE Yes Yes Yes Yes YesMaturity FE Yes Yes Yes Yes YesIssue Size FE Yes Yes Yes Yes Yes
Green Bonds 261 392 180 784 544Observations 32,194 48,952 21,354 100,174 65,457R2 0.647 0.718 0.753 0.824 0.726Adjusted R2 0.615 0.696 0.721 0.804 0.722
Table 8: Primary Market Analysis (sub-samples)The table shows results of the Fixed E�ects regressions of bonds Yield at Issuance in basis points forGreen and conventional bonds. The Issue Yield of Green and conventional �xed coupon plain vanillabonds is regressed against a Green indicator, which is 1 if the Bond is a green bond and 0 otherwise.Experienced is a dummy variable that is one if the bond is issued not on the same day as the �rst Greenbond of this issuer; Energy E�ciency, Renewable Energy, Clean Transport and Eligible GreenBond Project are dummy variables indicating the use of proceeds for the corresponding bond. Weinclude Issuer �xed e�ects, YearMonth �xed e�ects, Currency �xed e�ects, Rating, Seniority, Maturityand Issue size �xed e�ects, to take into account substantial di�erences between issuers, the yield curve,di�erent interest rate environments in di�erent countries and the in�uence of ratings on the yield atissuance. The Maturity �xed e�ects are three buckets for short-term (less than 5 years), medium (between5 and 10 years) and long term (more than 10 years) time to Maturity. The Issue Size �xed e�ects arethe deciles of the issue (in USD) in comparison to all other issues that occurred before or in the samemonth as the Issue.
In our second sub-sample analysis in Models (4) and (5), we consider green bond premia
18
Figure 6: Green bond premium over time
for di�erent types of issuers. After controlling for the di�erent �xed e�ects, we �nd
very high negative premia of around 31 bps for green bonds issued by governments and
supranationals, while the premia for corporate green bonds are much smaller or even not
signi�cantly di�erent from zero. Green bond investors, thus, rely more on the greenness
of bonds issued by o�cial entities.
In our �nal analysis with primary data, we investigate the variation of the premia over
time. To this end, we consider the coe�cient on the Green×YearMonth interaction
term in Model (3) in Table 7. We depict our result in Figure 6 starting from the end
of 2014 due to unavailability of su�cient data points before this time period. What we
�nd is that the premium is highly volatile over time. While there are months in a raw
with signi�cantly lower yields for green bonds during the hype period after the Paris
Agreement 2016 to 2017, in the last 2 years months with a non-zero Green coe�cient are
scarce and depend on single large issuances (e.g., government bonds issued by France in
January 2017 or Belgium in February 2018).
19
This result emphasizes the complexity of the pricing implications for green bonds as we
can not generalize the negative premium for all green bonds at all times just because of
the green label, and in particular, it again, underlines the importance of the credibility
of the issuance entity.
4.2 Results secondary market
Our results from the primary market analysis are based on a broader sample of green
and conventional bonds coming from companies issuing only green, only conventional or
both types of bonds. In this section we want to disentangle the green e�ect from other
possible e�ects on the bond price as much as possible. To this end, we focus only on
those green and conventional bonds, which are issued by companies with an experience
in both types of bonds. We have already seen in Table 6 that these bonds are usually
much larger than the average green or conventional bond in our total sample. Further,
the proportion of the Experienced green bonds, i.e., bonds issued not as the very �rst
green bond of this company, is higher for this sample (71% vs. 66%).
Having 777 and 3,504 bonds in our sample, we redo our regression analysis in equation
(1), where we now regress the Yield to Maturity on the Green dummy variable. Here
we also include a liquidity control (BidAsk), as this is particularly important on the
secondary market. We further include all �xed e�ects speci�ed in section 3.2 with the
exception of the Rating FE, which would signi�cantly reduce our sample.
Our main result for this analysis can be found in Table 9 in Model (1), while Models
(2)-(6) present our sub-sample analysis. In contrast to our main result using primary
data in Table 7 Model (3), where particularly strong demand by institutional investors
drives the result, we �nd the coe�cient on the Green variable to be signi�cantly posi-
tive for the full sample. On the secondary market, green bonds trade at 14 bps higher
yields than comparable conventional bonds, indicating that investors on these markets
have either lower demand for green investments or do not trust the green label in general
and have additional requirements on the issuer. Models (2) to (4) con�rm our results
20
Yield to Maturity
(1) (2) (3) (4) (5) (6)
Green 0.141∗∗∗ 0.251∗∗∗ 0.121∗∗∗ 0.224∗∗∗ 0.435∗∗∗ −0.020∗∗∗(0.003) (0.003) (0.004) (0.007) (0.006) (0.003)
BidAsk 0.312∗∗∗ 0.490∗∗∗ 0.294∗∗∗ 0.030 0.281∗∗∗ 0.428∗∗∗
(0.024) (0.019) (0.019) (0.044) (0.019) (0.049)
Subsample FULL EUR USD CNY CORP SOVR+SURPIssuer FE Yes Yes Yes Yes Yes YesYearMonth FE Yes Yes Yes Yes Yes YesCurrency FE Yes No No No Yes YesSeniority FE Yes Yes Yes Yes Yes YesMaturity FE Yes Yes Yes Yes Yes YesIssue Size FE Yes Yes Yes Yes Yes Yes
Green Bonds 777 191 304 49 408 369Conventional Bonds 3,504 1,136 1,171 163 1,724 1,780Observations 2,666,070 949,923 861,516 37,316 1,258,488 1,407,582R2 0.721 0.685 0.547 0.597 0.653 0.806Adjusted R2 0.721 0.685 0.547 0.595 0.653 0.806
Table 9: Secondary Market AnalysisThe table shows results of the Fixed E�ects regressions of bonds Yield to Maturity in basis points forgreen and conventional bonds. The Issue Yield of green and conventional �xed coupon plain vanillabonds is regressed against a Green indicator, which is 1 if the Bond is a green bond and 0 otherwise,and the corresponding BidAsk spread.
21
on the currency e�ects, with green bonds issued in USD having higher yields of 12 bps,
while bonds issued in CNY and EUR have on average 22-25 bps higher yields than their
conventional counterparts. Most interesting is the result for the corporate vs sovr/govt
sub-samples in Models (5) and (6). Here only green bonds issued by governments or
supranational entities are traded at a small but signi�cant premium of -2 bps, while cor-
porate bonds have on average 43 bps higher yields than conventional bonds. This �nding
clearly underlines the importance of the issuer's credibility for the greenness reputation
of the bond.
In our �nal analysis we investigate the determinants of the di�erence in yields between
green and conventional bonds in more detail. To this end, we compare mid yields of green
bonds with up to 10 comparable conventional bonds of the same issuer, thus, ruling out
all possible factors at the issuer level, which might have an impact on the corresponding
bond price.
The results of this regression analysis of the Green-minus-Conventional (GMC) vari-
able on di�erent controls are presented in Table 10. First, we observe a signi�cant positive
e�ect of the liquidity in terms of Bid-Ask spreads on the green bond premium. Across all
samples, a higher spread on green bonds is associated with a higher yield of up to 47 bps
for the green bond, and, thus, a lower price compared to the conventional one (except for
(Model (3) with CNY bonds). Further, we consider the e�ects of di�erences in coupons,
maturity and issuance sizes. Intuitively, the larger is the di�erence in coupons and ma-
turities between green and conventional bonds, the larger is the di�erence between the
corresponding yields to maturity. Further, the larger is the issuance size on the green
bond compared to the conventional one the lower is the yield required by the investors.
One of the interesting �ndings in this analysis is the e�ect of our Green exchange dummy
variable. Green bonds, listed on the exchanges with a dedicated green market segment,
such as for instance, in Luxembourg or London, trade at up to 8 bps (full sample in
Model (1)) and up to 13 bps (government and supranationals sub-sample in Model (5))
lower yields than their conventional counterparts. One possible explanation for this result
might be the higher visibility and transparency for those bonds provided through these
22
GMC
(1) (2) (3) (4) (5) (6) (7)
Constant 0.094∗∗∗ 0.076∗∗∗ 0.102∗∗∗ −0.220∗∗∗ 0.090∗∗∗ 0.115∗∗∗ 0.057∗∗∗
(0.002) (0.002) (0.003) (0.008) (0.002) (0.002) (0.002)
dBA 0.424∗∗∗ 0.465∗∗∗ 0.466∗∗∗ 0.044 0.455∗∗∗ 0.158∗∗ 0.336∗∗∗
(0.031) (0.013) (0.012) (0.029) (0.017) (0.066) (0.046)
greenEX −0.076∗∗∗ −0.039∗∗∗ −0.078∗∗∗ −0.133∗∗∗ 0.079∗∗∗ 0.017∗∗∗
(0.001) (0.002) (0.002) (0.001) (0.002) (0.002)
dcoupon 0.025∗∗∗ 0.032∗∗∗ 0.140∗∗∗ 0.131∗∗∗ −0.013∗∗∗ 0.076∗∗∗ 0.095∗∗∗
(0.001) (0.001) (0.003) (0.004) (0.001) (0.002) (0.002)
dmaturity 0.082∗∗∗ 0.126∗∗∗ 0.067∗∗∗ 0.085∗∗∗ 0.083∗∗∗ 0.080∗∗∗ 0.130∗∗∗
(0.000) (0.001) (0.000) (0.007) (0.000) (0.001) (0.001)
dsize −0.045∗∗∗ −0.035∗∗∗ −0.030∗∗∗ −0.007∗∗ −0.046∗∗∗ −0.034∗∗∗ −0.051∗∗∗(0.001) (0.000) (0.001) (0.003) (0.001) (0.001) (0.001)
ESGHigh 0.132∗∗∗
(0.003)
ESGLow 0.111∗∗∗
(0.004)
Subsample FULL EUR USD CNY SOVR+SURP CORP CORP
Green Bonds 649 186 210 33 344 305 165Pairs 4,609 1,439 1,581 229 2,560 2,049 1,096Observations 1,604,958 553,594 524,180 23,114 951,028 653,930 403,535R2 0.416 0.704 0.231 0.257 0.591 0.161 0.219Adjusted R2 0.416 0.704 0.231 0.257 0.591 0.161 0.219
Table 10: Determinants of the Green bond premiumThe dependent variable Green-minus-Conventional (GMC) is the mid yield spread between the pairsof green and conventional bonds. The dBA is the di�erence between the bid-ask-spread for green andconventional bonds; greenEX is an indicator which is one if and only if the green bond in the pair istraded at a green exchange; dcoupon is the di�erence in percent between the green and conventionalbonds coupon; dmaturity is the di�erence in years between the green and conventional bonds timeto maturity; dsize is the di�erence between the green and conventional bonds issue size; ESGHigh isan indicator which is one if and only if the issuers' combined ESG Score is within the top 30% of allissuers; ESGLow is an indicator which is one if and only if the issuers' combined ESG Score is withinthe bottom 30% of all issuers.
23
platforms. But also additional requirements from the exchanges such as a certi�cation
or third party opinion, or the alignment with one of the well-accepted Green Bond stan-
dards, increase the reliability of the green bonds market. However, the positive e�ect of
the Green exchange does not hold in Models (6)-(7), where we observe a positive coef-
�cient on the dummy variable, i.e., higher yields to maturity for green bonds, issued by
corporates. There might be di�erent explanations for this �nding. First, the ratio of cor-
porate green bonds listed on such exchanges is much lower than that of government and
supranational bonds (one third vs. more than one half). Second, the variation between
di�erent characteristics of corporate issuers is higher than that between o�cial entities.
There might be also other unobservable e�ects which have a higher impact on corporate
bonds. We can already see the reduction of the magnitude in the coe�cient by including
additional variables related to sustainability in Model (7). Here, we present our results for
a smaller set of green-conventional bond pairs (half of the sample) with available data on
ESG rankings. Interestingly, we �nd bonds issued by corporations within both, top and
bottom 30% in ESG ratings, to trade at higher yields of 11�13 bps than bonds, issued by
corporations in the middle 40%. The willingness of investors to pay for greenness of the
bond, thus, does not linearly increase with the ESG rating of the issuer, but seems to be
rather a parabolic function. A rationale explanation for this, at the �rst sight, unexpected
�nding might be that on the one hand, investors fear green�washing e�ects for bonds com-
ing from companies with a rather bad ESG reputation. On the other hand, they are not
willing to pay higher prices for green compared to conventional bonds issued by the same
issuer with a very good ESG reputation, since these companies in general spend money
on sustainable projects. The green label appears here to be rather a marketing name
without a fundamental change in the company's practices. A natural implication will be
that there might be no need for those companies to issue both types of bonds any more.
On the one hand, if the issuance costs for green bonds (certi�cation or reporting costs)
remain to be noticeably higher than those for conventional bonds, companies with top
ESG ratings will stop issue green bonds. On the other hand, with increasing liquidity,
established issuance processes and increasing demand and acceptance for green bonds on
24
the market, the demand for conventional bonds might decrease, in general. The result for
companies with very low sustainability ratings is more intuitive, as in particular, insti-
tutional investor apply top�down approach in their asset selection process, and specify,
thus, �rst the pool of suitable companies and then the corresponding instruments. This
result underlines the importance of the overall sustainability reputation of the issuer for
the acceptance of green bonds by investors.
25
5 Conclusion
With emerging green bonds market, the need for more transparency, uniform standards
regarding use of proceeds, certi�cation and reporting becomes more urgent. Many in-
vestors fear green�washing or green�labeling e�ects and particularly, private investors,
are skeptical about the sustainability hype. Though recent studies provide evidence that
investors generally value sustainability and are willing to pay for non�pecuniary charac-
teristics of investments, the existing results on so�called green bond premium are mixed.
In this paper, we provide an extensive and up-to-date analysis of the green bond premium
on di�erent markets and across di�erent investors and issuers. Using data on more than
1,500 green and 200,000 conventional bonds we reveal about 20�30 bps lower yields at
issuance for green bonds. This negative premium varies, however, across currencies and
issuer types and, particularly, over time. It is higher for bonds issued by governments
or supranationals, for bonds denominated in USD or EUR or bonds issued during the
hype years 2016-2017. For corporate bonds, the results are much weaker, indicating lower
institutional demand and higher issuance challenges for these bonds.
In our secondary market analysis, where we consider over 4,609 green�conventional bond
pairs of the same issuer, we �nd the di�erence in yields between green and conventional
bonds to be positive (i.e., lower prices for green bonds) in general. Only bonds, issued
by governments or supranational institutions trade at lower yields (-2 bps) than their
conventional counterparts. Further, we �nd the overall sustainability reputation of the
corporate issuer to be very important for the acceptance of green bonds by investors.
Here we observe positive premia for both, bonds issued by companies with a very low
and a very high ESG rating. A rationale explanation for this �nding might be that on
the one hand, investors fear green�washing e�ects for bonds coming from companies with
a rather bad ESG reputation, and green labeling e�ects for bonds issued by companies
with a very good sustainability reputation. Finally, we reveal the increasing importance
of the green bond exchanges as market catalysts, as bonds listed on exchanges with a
dedicated green market segment trade at signi�cantly lower yields.
26
References
Baker, M. P., D. B. Bergstresser, G. Serafeim, and J. A. Wurgler (2018):�Financing the Response to Climate Change: The Pricing and Ownership of U.S. GreenBonds,� SSRN Electronic Journal.
Barber, B. M., A. Morse, and A. Yasuda (2018): �Impact Investing,� SSRN Elec-
tronic Journal.
Ehlers, T. and F. Packer (2017): �Green bond �nance and certi�cation,� BIS Quar-
terly Review.
Flaherty, M., A. Gevorkyan, S. Radpour, and W. Semmler (2017): �Financingclimate policies through climate bonds � A three stage model and empirics,� Researchin International Business and Finance, 42, 468�479.
Flammer, C. (2018): �Corporate Green Bonds,� SSRN Electronic Journal.
Glomsrød, S. and T. Wei (2018): �Business as unusual: The implications of fossildivestment and green bonds for �nancial �ows, economic growth and energy market,�Energy for Sustainable Development, 44, 1�10.
Hachenberg, B. and D. Schiereck (2018): �Are green bonds priced di�erently fromconventional bonds?� Journal of Asset Management, 19, 371�383.
Hartzmark, S. M. and A. B. Sussman (2019): �Do Investors Value Sustainability?A Natural Experiment Examining Ranking and Fund Flows,� Journal of Finance,forthcoming.
Helwege, J., J.-Z. Huang, and Y. Wang (2014): �Liquidity e�ects in corporate bondspreads,� Journal of Banking & Finance, 45, 105�116.
Karpf, A. and A. Mandel (2017): �Does it Pay to Be Green?� SSRN Electronic
Journal.
��� (2018): �The changing value of the `green' label on the US municipal bond mar-ket,� Nature Climate Change, 8, 161�165.
Nanayakkara, M. and S. Colombage (2018): �Do Investors in Green Bond MarketPay a Risk Premium? Global Evidence,� SSRN Electronic Journal.
Preclaw, R. and A. Bakshi (2015): �The Cost of Being Green,� Tech. rep., Barclaysresearch.
Reed, P., T. Cort, and L. Yonavjak (2019): �Data-Driven Green Bond Ratings asa Market Catalyst,� Journal of Investing, 28, 66�76.
Zerbib, O. D. (2019): �The e�ect of pro-environmental preferences on bond prices:Evidence from green Bonds,� Journal of Banking & Finance, 98, 39�60.
27