PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Green finance links investment to firms' environmental performance. In the past few years, for example, green bonds drove rapidly increasing amounts to firms' climate-friendly projects, and the existing empirical literature suggests that it works: When firms issue green bonds, their stock price rises, and their CO2 emissions decrease subsequently. To explain these facts, much research has looked for a green bonds' yield spread, but investors' concern for the environment does not appear to currently play a significant role. By contrast, we suggest that (1) firms' managers use green bonds to signal the economic efficiency of their green projects to investors, and that (2) they do so because they are concerned about their firm's stock price, unlike the traditional view of corporate social responsibility. Our model predicts a relationship between firms' proportion of green bonds, managerial incentives, and carbon pricing. We test this prediction by exploiting both cross-industry differences in the stock price sensitivity of managers' pay and in share turnover, and cross-country variations in effective carbon prices. Our results not only support the role that our theory ascribes to managerial incentives, but also show that this role positively depends on carbon pricing, whose effect is, therefore, amplified by green bonds. JEL classification: D53; H23; G14; Q54.
Content may be subject to copyright.
Why Do Firms Issue Green Bonds??
Julien Xavier Daubanes
University of Geneva (GSEM), MIT (CEEPR), and CESifo
E-mail address:
Shema Fed´eric Mitali
Ecole Polytechnique F´ed´erale de Lausanne (CDM)
E-mail address:
Jean-Charles Rochet
University of Geneva (GSEM), Swiss Finance Institute, and
Toulouse School of Economics
E-mail address:
November, 2021
?. We thank Tom Steffen for excellent research assistance. We have benefited from useful comments
by participants at various seminars and conferences: Toulouse Business School; University of Geneva;
University of Verona; McGill; University of Montpellier; OECD; Congress of the SSES in Geneva; SAET
Conference in Ischia; FAERE Conference in Rennes; University of Gothenburg; TSE Conference on
Markets, Morality, and Social Responsibility; Geneva Graduate Institute; PSE; EAERE Conference;
CESifo Conference on Energy and Climate Economics; University of Nantes; AERE Summer Conference;
Bank of France; GRASFI Conference. Particular thanks to Diego Cardoso, Pierre-Andr´e Chiappori,
Antoine Dechezleprˆetre, Mathias Dewatripont, Ivar Ekeland, Andr´e Grimaud, Henry Jacoby, Rapha¨el
Levy, Tom Lyon, Rick van der Ploeg, Brittany Tarufelli, and Jean Tirole. The research leading to these
results has been supported by the Swiss National Science Foundation (SNSF) within the framework of
the National Research Programme “Sustainable Economy: resource-friendly, future-oriented, innovative”
(NRP 73).
Green bonds allow firms to commit to climate-friendly projects. Shareholders react pos-
itively to their announcement. Based on prior empirical studies, we suggest that green
bond commitments help managers signal the profitability of their green projects and that
they do so because they are sensitive to their firm’s stock price. We present a signaling
model in which firms undertake green projects not only because of carbon penalties but,
additionally, because of managerial incentives, predicting that the role of the former is
augmented by the latter. We test this prediction by exploiting both cross-industry dif-
ferences in the stock-price sensitivity of managers’ pay and in stock share turnover, and
cross-country variations in effective carbon prices. Our results not only support the role
that our theory ascribes to managerial incentives, but also show that this role mainly
depends on carbon pricing. With green bonds, governments do not avoid carbon pricing.
On the contrary, the latter is essential to the effectiveness of the former.
JEL classification: D53; H23; G14; Q54.
Keywords: Green bonds; Green finance; Climate policy; Carbon pricing; Managerial
incentives; Short-termism.
I. Introduction
Green finance certification allows investors to link their decisions to firms’ commit-
ments towards the environment. Green bonds are the most emblematic and prominent
green finance instrument: Their issuers commit to use the bond proceeds to a certified
climate-friendly project. For example, Unilever announced on March 19, 2014, one of
the now most famous certified green bond issues, earmarking more than 400m to new
climate-friendly production capacities. Namely, this commitment confirmed the success
of various years-long research plans—see Unilever’s 2010 “Sustainable Living Plan”—to
develop new technologies and products, such as climate friendly detergents and refriger-
ants. Investors enthusiastically received the announcement, generating stock returns1of
more than 5% over a window of [5,+5] days around the announcement date. Similarly,
Apple, for example, issued more than 4.5bn of certified green bonds between 2016 and
2019, to develop its use of clean energy sources and to improve its energy efficiency. In
the past few years, a rapidly increasing number of firms have made similar commitments,
leading to a boom in the global green bond market, whose volume has nearly doubled
every year since 2013—see the amounts of green bonds issued in Figure I—expanding to
91.1bn in 2021, around 4% of total corporate bond issuance (Central Banking, 2021).
In this paper, we build on prior empirical studies of the green bond boom to present a
theory for why firms commit to CO2 reducing projects through certified green bonds. We
give an explanation for green bond issuance that relies on managerial incentives, and we
empirically validate the role of the former, with implications for the relationship between
green bonds and public policies.
Firms’ issuance of green bonds is voluntary,2but seem nevertheless environmentally
1. See Section V for details on our event-study estimation of abnormal stock returns.
2. Certified green bonds must finance projects that satisfy some standards as, for example, the Cli-
mate Bond Standards or the Green Bond Principles. In particular, these standards provide a verification
scheme. Compliance with green bond commitments is also voluntary, but noncompliance seems costly.
There are only few examples of “green defaults,” probably because noncompliance causes a significant
reputational loss. For instance, when Repsol’s 500m green bond, initially certified, was finally deemed
noncompliant with the Climate Bonds Standards, it was excluded from green indexes. This also sug-
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
Amount of green bonds issued by year and type
Figure I: Green bonds issuance
effective. Flammer (2021) shows that firms issuing certified green bonds largely reduce
their CO2 emissions subsequently3and argues against the possibility of greenwashing.
Yet very little is known about the mechanisms that make green bonds work.
In the face of the climate problem, economists keep recommending to price carbon
(e.g., Stavins, 2008). In practice, however, this direct approach is less successful than
hoped.4Moreover, alternatives either to publicly support climate-friendly initiatives
or to restrict technologies are undermined by governments’ indebtedness, and limited
information and expertise. The urgency of the climate challenge, therefore, calls for
examining all instruments that are feasible and potentially effective. This need is reflected
by recent studies on the effectiveness of other options such as, for example, energy-
efficiency standards (Jacobsen, Knittel, Sallee, and van Benthem, 2021) or environmental
disclosure policies (e.g., Krueger, Sautner, Tang, and Zhong, 2021). As a matter of fact,
gests that current certification standards are relatively consensual among investors, despite controversies
(Environmental Finance, 2017).
3. Using a matching methodology, Flammer estimates that firms issuing certified green bonds reduce
their CO2 emissions by 13% over the course of the next two years; not all of this reduction may be
attributed to projects financed through green bonds.
4. Even in developed countries, effective carbon prices are far below the social cost of carbon emissions
(OECD, 2018).
the rapid growth of the green bond market is receiving a lot of attention by governments
and financial institutions.5But green finance instruments lack theoretical and empirical
foundations. This paper is a first attempt to fill this gap.
Recent empirical analyses of the green bond boom establish the following stylized
facts. First, firms’ stock price increases when they announce the issue of certified green
bonds and financed projects. Tang and Zang (2018), Baulkaran (2019), and Flammer
(2021) find relatively large abnormal stock returns of 0.5-1.5% around the announce-
ment of certified green bonds. This is reminiscent of environmental awards (Klassen and
McLaughlin, 1996) and unlike conventional bonds (Eckbo, 1986; Mikkelson and Partch,
1986; Antweiler and Frank, 2006). Second, firms’ certified green bonds do not allow them
to obtain less costly financing. Empirical estimates of the green bond yield spread are
very low, ranging from 0 to 0.2% (Tang and Zang, 2018; Zerbib, 2019; Flammer, 2021;
Kapraun and Scheins, 2020).6This—along with qualitative evidence of industry practice
(Chiang, 2017)—indicates that concerned investors do not currently play a significant
role (Harrison, Partridge, and Tripathy, 2020).7Neither does bond default: The absence
of green bond yield spread means that green bonds do not significantly affect firms’ re-
payment capacity. Third, certification of green bonds is critical. So-called “self-labeled”
green bonds are associated with neither CO2 reduction, nor stock market reaction, nor
bond yield spread (e.g., Flammer, 2021).
We suggest that firms’ managers issue certified green bonds because the latter signal
the profitability of green projects to investors and that managers do so because they are
5. “Over the last few years, the ECB raised the share of green bonds in its own-funds portfolio to
3.5% in 2020, while planning to further increase it in the immediate future.” (Central Banking, 2021)
6. As Marilyn Ceci, Head of Green Bonds at JP Morgan, sums up, green bonds “generally price in
line with traditional bonds, but occasionally demand outstrips supply and they can price a few basis
points tighter” (Harrison, Partridge, and Tripathy, 2020).
7. Some investors certainly have a preference for green bonds and for the firms issuing them (Flammer,
2021). Yet the absence of spread means that the currently marginal investor is not significantly concerned.
Pastor, Stambaugh, and Taylor (2021) show that investors’ concern increases as climate change become
more manifest. Green investors are likely to take a more active part and a green bond spread may
emerge—see, for example, Financial Times, January 4, 2021, “Analysts expect as much as 500bn of
green bonds in bumper 2021”. Our theory accommodates concerned investors and a green bond spread.
See Appendix C.
sensitive to their firms’ stock price. More precisely, we combine two main ingredients.
First, we present a model in which green bonds are a signaling device, conveying positive,
although imperfect, information about the expected profitability of their environmental
investments. Stock investors find it more difficult to assess the profitability of green
projects with respect to business-as-usual activities. And only firms with the most prof-
itable green projects would commit to undertake them. It is the information that green
bonds reveal that explains abnormal announcement stock returns. The informational role
of green bonds is supported by Tang and Zang’s (2018) finding that stock markets react
mainly to the first financing of green projects and much less to their refinancing.8
Second, our model features managers’ interest in the stock price of their firm. If man-
agers only cared about future profits, signaling would be useless. Managerial concern
for short-term returns, also sometimes coined “short-termism,” has various origins. One
is that managers’ actual compensation schemes include stock components (Stein, 1989;
Georgen and Renneboog, 2011). For example, Edmans, Gabaix, and Landier (2009) mea-
sure the sensitivity of managers’ compensation to their firms’ stock price;9and Gopalan,
Milbourn, Song, and Thakor (2014) show that this is mainly a short- to medium-run sen-
sitivity.10 Besides their compensation, managers’ short-term incentives result from risks
of takeover (Stein, 1988), short-term investors (Bolton, Scheinkman, and Xiong, 2006),
and markets’ attention to short-term returns (Summers and Summers, 1989). Summers
and Summers (1989) suggest, and Cremers, Pareek, and Sautner (2020) confirm, that
investors’ short-termism and managerial myopia are reflected by stock share turnover.
Cross-industry variations in both managerial compensation sensitivity to the stock price
and share turnover are significant. Our theory will focus on these variations, which we
will also exploit empirically.
8. Moreover, Flammer (2021) finds that these reactions vanish when green bonds are issued by finan-
cial intermediaries.
9. More precisely, their wealth-performance sensitivity variable is the dollar change in the CEO’s
wealth for a 100 percentage point change in the stock price, scaled by annual pay. This measure is
independent of firm size and is thus comparable across firms of different size.
10. They find that the vesting period of most of executives’ stock and options grants is less than five
years, to be compared with the average maturity of 7.5 years of the certified green bonds in our sample.
Besides green bonds, public policies in most countries already provide firms with
some, although insufficient, incentives to undertake CO2 reducing projects (OECD, 2018):
carbon tax, if any, excise taxes on carbon energy sources, or emission trading schemes.
In our model, these policy-induced incentives are captured by an exogenous carbon price.
With green bonds, the effect of carbon pricing is twofold: It induces firms to undertake
more certified green projects not only because it penalizes more conventional technologies,
but also because, all else unchanged, it amplifies the stock market reaction to green bonds
and, therefore, managers’ interest in certified green projects.
We derive a testable relationship between the proportion of green bonds issued by an
industry, the carbon price this industry is applied, and managers’ concern for their firms’
stock price. In this relationship, the positive effect of managers’ sensitivity to the stock
price on green bonds captures the role that our theory ascribes to managerial incentives.
This positive effect is due to the interaction between managerial incentives and carbon
We use data that relate public firms’ certified green bonds to their environmental
and financial characteristics, and to the effective carbon price that prevails where they
are based. First, we exploit carbon price variations across countries and industry-level
variations of the stock price sensitivity of managers’ compensation to validate the role of
managerial incentives in green bond issuance. This has implications for carbon pricing,
whose impact is amplified by green bonds; but also for green bonds, whose effectiveness
relies on carbon prices. Second, we exploit the recent implementation of green bond
policy support in various countries to corroborate existing evidence that green bonds can
be considered as CO2 reduction commitments.
Environmental economists often point out that well-intentioned initiatives need not
be effective. Green bonds may be the opposite. They seem environmentally effective, but
their success is likely to be mainly a matter of financial interest and short-term incentives.
Our study is at the intersection of the above finance literature on managerial compen-
sation and the literature on the private and social benefits of firms’ socially responsible
initiatives. enabou and Tirole (2010) hold that firms’ corporate social responsibility
reflects both the mitigation of excessive managerial short-termism and the expression of
stakeholders and managers’ concern about unresolved external effects. Maxwell, Lyon,
and Hackett (2000) point to firms’ interest in deterring future political action. Heal
(2005) and Daubanes and Rochet (2019) further stress that self-regulation avoids costly
conflicts with the rest of society. These rationales for responsible business conduct relies
on the idea that it benefits firms in the future so that it should be enhanced by manage-
rial long-termism. Similarly, the frameworks of analysis of Magill, Quinzii, and Rochet
(2015), and Hart and Zingales (2017), as well as Edmans (2020) imply that firms should
adopt a more inclusive perspective. Perhaps surprisingly, our results suggest that the
recent boom in green finance, although apparently effective environmentally, has mainly
to do with investors’ financial interest and managers’ short-term incentives. We stress,
moreover, that firms’ voluntary green finance commitments are complementary to public
Our theory focuses on a minimal set of ingredients. We model an industry in which
each firm considers a single incremental project over two periods of time. In the first
period, firms’ managers decide whether to undertake their project in a conventional or in
a green (CO2 reducing) fashion. We take as given the sensitivity of their objective to stock
prices. Nor do we address the imperfections of environmental certification. Instead, we
focus on the mechanisms underlying its effectiveness. We consider that green projects are
all financed through green bonds, which perfectly certify the adopted green technology,
while other projects are financed by conventional securities, among which we do not
make any distinction. Shareholders do not observe the profitability of green projects,
but infer it from managers’ commitments through green bonds and the carbon price in
effect. They price firms’ stock accordingly in the first period, anticipating rationally the
profits realized in the second period. Finally, since investors’ concern for the environment
currently seems of second-order practical importance, we purposely omit this aspect in
the baseline version of our model. However, we provide an extension of our analysis
to the presence of environmentally concerned investors, the emergence of a green bond
yield spread, and an ESG-augmented stock market reaction, with no implications for the
validation of our theory.
The rest of the article is structured as follows. In Section II, we relate our paper to
various economic literatures. In Section III, we present our baseline model of green bonds,
and derive our main testable prediction. In Section IV, we describe our data samples,
and use them to test the main prediction of our model. Section V provides additional
empirical findings that support our theory. In Section VI, we discuss the case of concerned
investors—with technical details in the Appendix—and draw policy implications.
II. More of the related literature
In addition to the studies of corporate social responsibility and to the empirical studies
on green finance, our findings are related to literatures on certification, climate policy
instruments, managerial incentives, and aspects of new green finance markets.
Certification.—The literature on financial certification—e.g., Farhi, Lerner, and Tirole
(2005), and Lerner and Tirole (2006)—and on credence goods’ labelling for consumers—
e.g., among many others, Bonroy and Constantatos (2015), or Bonneton (2020)—focuses
on situations in which agents are directly interested in the certified information: pro-
duction’s environmental impact, the potential of an innovation, credit risk, etc. In this
context, Lyon and Fisher (2014) and Bouvard and Levy (2018) examine how certifiers
set standards’ stringency and accuracy.
In our baseline model, by contrast, investors are not directly interested in the certified
environmental performance, but only indirectly because firms’ financing commitments
reveal economic efficiency. Moreover, we overlook the choice of standards’ stringency so
as to focus our analysis on the mechanism that makes green finance effective in practice.
Climate policy instruments.—Much empirical research effort has been devoted to
the pigovian solution to the carbon externality—see, e.g., Nordhaus (2017). In the
face of large remaining carbon pricing gaps (OECD, 2018), some examine second-best
instruments—see, among other examples, Jacobsen et al. (2020) on energy efficiency
standards—and how these instruments complement some carbon pricing (e.g., Dimanchev
and Knittel, 2020). Others have examined the Coasian type of voluntary arrangements
and the fundamental role of information disclosure (Tietenberg, 1988). Voluntary actions
raise the question of their effectiveness. Khanna and Damon (1999) suggest that volun-
tary programs might be effective because of significant public recognition efforts targeting
customer goodwill. Denicol`o (2008) suggests that firms’ voluntary environmental actions
may seek to obtain excessively stringent regulation at their benefit.
Our analysis is complementary to this literature. Green bond certification specifically
targets investors. We point to a new mechanism through which voluntary green finance
commitments can effectively complement carbon pricing.
Managerial incentives.—Our theory relies on managers’ sensitivity to the stock price
of their firm. Our analysis takes an agnostic perspective on the causes of managerial
concern for their stock price. The finance and business literatures have attributed such
managerial incentives to: the types or preferences of shareholders (e.g., Polsky and Lund,
2013; Bebchuk, 2021); short-term investors (Summers and Summers, 1989), and result-
ing options for shareholders (Bolton et al., 2006); the need to incentivize managerial
efforts (Marinovic and Varas, 2019), especially in a context of competition for executives
(Thanassoulis, 2013); the horizon of industry projects and their mispricing (Schleifer and
Vishny, 1990); the risk of takeover (Stein, 1988).
Managers’ incentives are reflected in their compensation schemes, which are based on
stock prices, through shares and options mostly exercisable over relatively short periods
(Gopalan et al., 2014). For example, Edmans et al. (2009) measure the sensitivity of
managers’ compensation value to the stock price. Our empirical analysis will make use
of this measure and its variations across industries, confirming its practical relevance.
Some have pointed out that managers’ short-term incentives have various, mostly
detrimental, effects: information manipulation to inflate current profits (Sobel, 1985;
enabou and Laroque, 1992); financial instability (Summers and Summers, 1989); (lack
of) long-term information acquisition (Casamatta and Pouget, 2015); (lack of) long-term
investments (Ladika and Sautner, 2020; Cremers, et al., 2020); (lack of) corporate social
responsibility (B´enabou and Tirole, 2010). With respect to these studies, we suggest that
managerial incentives may also motivate certified green finance commitments.
Green finance.—This paper adds to a burgeoning literature that examines various
other aspects of the recent development of green finance: shareholder activism (Gollier
and Pouget, 2009); the selection of green projects (Kotchen and Costello, 2017); divest-
ment (Chava, 2014); stock investors’ concern (Gibson Brandon, Kr¨uuger, and Mitali,
2020; Pastor at el., 2021); the impact of concerned investors (Landier and Lovo, 2021);
municipal green bonds (Baker, Bergstresser, Serafeim, and Wurgler, 2020); climate risks
(e.g., Barrage and Furst, 2019); stranded assets (e.g., van der Ploeg and Rezai, 2019).
III. A simple theory of green bond issuance
In the model of this section, we consider that investors care only about their invest-
ments’ financial returns, and not at all about these investments’ environmental impact.
Proofs are relegated to Appendix A. Appendix B presents a special case that yields
explicit solutions. Section VI will discuss, with technical details in Appendix C, how
this model accommodates the case of investors who are sensitive to firms’ environmental
III.A. Technology
The industry consists of a continuum of firms. Firms have regular activities and
incremental projects, a single one per firm. We assume that firms’ regular activities
are immutable. Their incremental projects, however, may be green or not, as we now
Firms’ projects take place over two dates t= 0,1. All projects require one unit of
capital at date t= 0 and generate the same revenue Y > 0 at t= 1. As will be clear
shortly below, projects will differ in profitability because of their cost.11 There are two
technological options: At date t= 0, firms’ managers choose whether their project will
be green or conventional (brown), which will be indicated by subscripts k=Gand k=B
Projects’ technology k=G, B determines their environmental performance. At date
t= 1, green and conventional projects generate CO2 emissions xGand xBrespectively,
with xB> xG0. Emissions are penalized at an exogenous rate τ0.12 Although, in
reality, large firms’ total CO2 emissions are often estimated, they can hardly be attributed
to individual projects unless these projects are certified. We assume, therefore, that
emissions due to firms’ projects are not directly observable.
Green and conventional projects differ by their cost and by investors’ ability to assess
this cost. All projects entail a business-as-usual cost cB>0 that is perfectly known. As a
matter of fact, green technological options are less usual. We assume that green projects
further entail an additional cost ∆c(i)—maybe negative—strictly increasing, and convex
function of the type of firms’ incremental project i[0,1], that is private information
of firm i’s manager. The type of firms’ projects ranks them in decreasing order of their
green efficiency.
To sum up, the cost of firm i’s project is:
ck(i) = cBif k=B
cB+ ∆c(i) if k=G, i [0,1] .
The resulting industry’s marginal cost of emission reduction is the increasing function
c(i)/x, where ∆xxBxBis the project-level potential CO2 reduction. Note that
this cost may be interpreted broadly, as net of side benefits for the firm to develop green
projects, such as learning-by-doing, marketing, etc.
11. This is without loss of generality. Projects could be equivalently assumed to differ in their revenue
rather than their cost.
12. In practice, effective carbon prices include not only explicit carbon penalties such as carbon taxes
and allowance prices, if any, but also specific excise taxes on carbon containing energy inputs.
III.B. Green finance and firms’ problem
We assume that all incremental projects are financed by bonds and that green bonds
perfectly certify the green technology of financed projects.13 We only consider certified
green bonds, i.e., we ignore so-called self-labeled green bonds, because empirical evidence
shows that the latter not only are environmentally ineffective but also do not trigger
investors’ reaction (e.g., Flammer, 2021).
As a matter of fact, green and conventional bonds have similar returns—see, e.g., Tang
and Zang (2018) and the literature reviewed by Harrison et al. (2020). In the baseline
model of this section, we assume that both types of bonds repay the same14 R1 + r
at date t= 1, where ris the risk-free interest rate, which we take as exogenous.
To sum up, an incremental project of type i[0,1] with technology k=G, B
generates an additional profit at date t= 1:
πk(i) = YRck(i)τxk+εk(i),(1)
where εk(i) is a technology-specific random term with E[εk(i)] = 0, k=G, B, and where
Besides their incremental project, firms may also differ by the profits generated by
their regular activities at date t= 1: V+ε, where V0 is the known firm-specific
expected profit, εis a random term with E[ε] = 0, and both are independent of projects’
type. Profits (1) from firms’ incremental project add to profits generated by their regular
Firms’ managers correctly anticipate expected future profits V+E[πk(i)]. By contrast,
investors do not observe the type of firms’ project, but to the extent that certified financial
decisions partly reveal the efficiency of green projects. Firms’ date-0 stock prices Skwill
depend on their project’s certified technology k=G, B, as will be determined further
13. Conventional bonds, including self-labelled, non-certified green bonds may be interpreted as regular
bank loans.
14. In Section VI, we describe—with technical details in the Appendix—how this model easily accom-
modates concerned investors and a yield spread between green and conventional bonds.
Managers care not only about firms’ future expected profits V+E[πk(i)], but also
about their current stock price Sk. The finance literature points to various sources of
short-term managerial incentives. On the one hand, they partly originate from share-
holders’ own interest in—or need to rely on—their current stock price, as is reflected
by the actual structure of executives’ compensation, with shares and options with rela-
tively short vesting periods. On the other hand, besides compensation-related incentives,
managers are concerned about the risk of takeover (Stein, 1988) and the attention that
markets pay to short-term performance (Summers and Summers, 1989).
Our theory is agnostic about the origins of managers’ concern for their stock price.
We adopt Stein’s (1989) direct modeling of managers’ objective. Managers choose their
project’s technology k=G, B in such a way as to maximize:
Uk(i) = (1 α)V+E[πk(i)]
1 + ρ+αSk,(2)
where ρis the discount rate and the parameter α(0,1) captures the stock price
sensitivity of managers’ compensation that prevails in the industry, which we both take
as given.
III.C. Timing
The timing is represented in Figure III. Prior to date t= 0, firms’ projects are
indistinguishable and the ex ante stock price Sonly differs across firms because of their
regular profits V.
At date t= 0, managers choose to undertake their incremental projects in a green
(k=G) or conventional way (k=B). Green projects are certified and financed through
green bonds. At the same time, stock prices become SGand SBfor firms undertaking
green and conventional projects respectively.
At date t= 1, firms’ regular activity takes place and their project is realized under
the committed technology, bonds are repaid to bond investors, and the resulting profits
accrue to shareholders.
t < 0Ex ante stock market equilibrium
Green projects’ certification
Stock market reaction
Bonds’ repayment to bond investors
Profits to shareholders
t= 0
t= 1
Figure II: Timing
III.D. Green bond supply
Managers take as given firms’ stock prices SGand SBaccording to whether they under-
take their incremental project in a green or conventional way. For simplicity, throughout
this section, we will express stock market reactions in terms of stock prices:
S ≡ SG− SB.(3)
In Section V, we will show that these reactions can equivalently be expressed in terms
of stock returns, and will estimate “abnormal stock returns.” For now, these reactions
are parametric; equilibrium stock values and returns will be endogenously determined
further below.
Despite its rapid growth in the past few years, industries’ green bonds only represent
4% of their volume of debt. We will, therefore, focus the analysis on interior equilibria
in which 0 < ie<1. This rules out unrealistic situations in which the effective carbon
price is either extremely high or extremely low.15
In particular, our setting assumes that E[πB]0, so the carbon price is not sufficiently
high to discourage firms to undertake their projects. Indeed, we are mainly interested
in firms’ adoption of green technologies, less so in their decision to implement or not
conventional projects. In general, voluntary approaches do not affect the latter margin—
see, for example, Lyon and Maxwell (2003) in the context of voluntary agreements.
Appendix A shows how managers choose to make a green project in a way that
balances the additional cost of green technologies and their benefit of the stock market
reaction to green certification.
Lemma 1 (Green bond supply). Given SGand SB, a firm with a project of type i
chooses a green technology k=Gif and only if iie, where ieis characterized by:
(1 α) (∆c(ie)τx) = α(1 + ρ)∆S,(4)
where S ≡ SG− SB.
The proportion of green projects ieincreases with the stock market reaction S.
For a given stock price reaction to green bond issuance ∆S, equation (4) determines
the supply of certified green projects ie, as depicted in Figure III in which green projects
are types i[0, ie]. The relationship is increasing: As the stock reaction ∆Sis more
pronounced, managers are willing to undertake and certify more green projects. Moreover,
all else held unchanged, the supply of green bonds is increased both by the carbon price τ,
which penalizes more the profit of conventional projects, and by the degree of short-term
incentives α, which gives more weight to the stock price reaction.
The next step of the resolution is to derive the stock price reaction to green bonds.
If there were no green finance certification of firms’ technological choice, stock prices
would not adjust at all.16 Then, the equilibrium proportion of green projects i0would be
15. Conditions under which the equilibrium proportion of green bonds in the industry is interior are
c(0) < τx < R1
0c(i)di and τxBYRcB—see Appendix A for details.
16. In Section V, Table II indicates that self-labelled green bonds, unlike certified ones, do not generate
abnormal returns that are statistically different from zero.
determined by the standard equality between marginal cost and the incremental carbon
c(i0) = τx. (5)
In Figure III, this equilibrium is depicted at the intersection of the rising green bond
supply curve with ∆S= 0.
With certification, however, we will show further below that SG>SBin equilibrium,
so that green bonds induce managers to undertake more green projects than they would
otherwise. Green bond certification does not require “additionality,” i.e., green bonds also
finance projects that would have been undertaken without green bonds. In our baseline
model, all additional green projects are due to the positive stock market reaction to green
III.E. Stock market reaction to green bonds
At date t= 0, investors take as given firms’ commitment to their green projects, and
use it to both infer the profitability of these projects and assess date-1 firms’ value:
1 + ρ, V 0, i [0,1], k =G, B. (6)
Prior to date-0 firms’ announcements, investors rationally anticipate date-0 equilibrium
proportion of green projects, so ex ante stock prices are:17
S=ieSG+ (1 ie)SB.(8)
Appendix A shows the following results.
17. Consequently, ex ante stock prices may also be written:
1 + ρ,(7)
which depends not only on firms’ specific regular expected profit Vbut also on the given proportion of
green projects ie. In our model, Swould be maximum if the anticipated iewas equal to i0, the equilibrium
proportion of green projects that would prevail if managers were focused on long-run profits (α= 0).
In the neighborhood of the green finance equilibrium in which ie> i0,Sis decreasing with iebecause
a greater proportion of green projects leads the industry further away from profit maximization—see
Appendix A for details.
Lemma 2 (Stock market equilibrium). Given the proportion ieof green bonds, the
stock price reaction to firms’ green bond issuance is:
(1 + ρ)∆Se=τxE[∆c(i)|iie].(9)
This reaction becomes less pronounced as the proportion of green bonds iebecomes
The stock market reaction to green bonds stems from the information that green bond
commitments convey about the profitability of green projects. In our model, prior to
firms’ financing decisions, investors cannot make any distinction between firms’ projects.
With the announcement of green bond financed projects, they reassess that these projects
are more profitable than expected, iie, both because these projects will be relatively
favored by carbon pricing and because their expected costs become E[∆c(i)|iie] rather
than the unconditional E[∆c(i)].
Formula (9) tells that the magnitude of the stock market reaction to green bonds is
determined by the difference between the avoided carbon penalty and investors’ expected
additional cost—benefit, if negative—of green projects. The former will translate into
an amplification effect of carbon pricing: Due to green bonds, carbon pricing increases
investors’ reaction to green commitments, an effect that would vanish without green
The expected additional cost of green projects—benefit, if negative—increases with ie
because more green bonds mean that less efficient green projects are undertaken. There-
fore, whether it contributes to the stock market reaction negatively or positively, it will
translate into a dilution effect of green bonds.
When ∆Seis positive, the stock price of firms issuing green bonds increases at date
t= 0, so green bonds benefit in the short term the shareholders of firms that issue them,
whether financed green projects would have been undertaken absent green certification
(0 ii0), or whether these green projects are additional (i0< i ie). For additional
projects, however, this short-term benefit comes at the cost of reduced profits at date
t= 1.
III.F. Equilibrium proportion of certified green projects
In the rational expectation equilibrium, the stock market reaction characterized in
(9) is consistent with the supply of green bonds in (4) that this reaction generates. We
examine the resulting equilibrium proportion of green bonds in the industry, and its
determinants. We obtain following results.
Proposition 1 (Green bond rational-expectation equilibrium).
1. The equilibrium with rational expectations exists and is unique;
2. In this equilibrium:
(a) The stock market reaction to green bonds is positive: Se>0;
(b) The resulting proportion of green bonds increases with both the industry’s
managerial stock-price sensitivity αand the carbon price τ;
The intersection of relationships (4) and (9) determines the unique equilibrium pro-
portion, ie=ie(α, τ ), as depicted in Figure III. It depends on both managerial incentives
and carbon pricing, whose joint role will be empirically exploited in the next section.
In this equilibrium, the stock market reaction to green bonds is always positive:
Se=(1 α) (∆c(ie)E[∆c(i)|iie])
1 + ρ>0.(10)
Managers do not issue green bonds as much as to make the positive stock reaction vanish.
In Section V, we show that (10) translates into positive stock returns that depends on the
relative size of projects. We also provide event-study estimates of stock returns around
the announcement of green bonds and find that they are both positive, and higher when
green bonds are large with respect to firms’ size.
α(1 + ρ)∆S
α(1 + ρ)∆Se
(1 α) (∆c(i)τx)
Figure III: Equilibrium proportion of green projects
The resulting proportion of green bonds is determined by the equality between the
incremental carbon penalty and the marginal cost of green projects where the latter is
adjusted by managers’ benefit from green bonds’ signal:
(1 α)∆c(ie) + αE[∆c(i)|iie] = τx= ∆c(i0),(11)
where ∆c(ie)>E[∆c(i)|iie] implies both that the equilibrium proportion of green
bonds is higher than the proportion of green projects in absence of certification (ie> i0),
and that the former increases with managers’ stock-price sensitivity α.
The positive effect of managerial stock-price sensitivity αon green bond issuance is
the main implication of our theory. Our analysis explains that this is driven by the stock
price reaction to green bonds, and that this reaction, all else held unchanged, is more
pronounced as the effective carbon price is higher.
In general, our model characterizes the role of αand τin ie(α, τ ) implicitly, as we just
described, and as one can show by shifting curves of Figure III. Linear testable impli-
cations requires a linear approximation. In fact, assuming that the marginal abatement
cost function ∆c(i) is affine is sufficient to obtain a directly testable explicit expression
of the proportion of green bonds.
Corollary 1 (Linear testable prediction). Assume that the green technology cost is
affine: c(i)a+bi, with b > 0. Redefine managers’ stock-price sensitivity as
The equilibrium proportion of green bonds in the industry takes the closed-form
linear expression:
| {z }
b˜ατ a
| {z }
managerial incentives
To test our theory, in Section IV, we will attempt to estimate the linear expression
In this expression, the first part reflects the proportion of green projects that would
be undertaken in the absence of certification: It only depends on τand is independent of
managerial incentives.
The rest of the left-hand side of (12) is positive and reflects the role of managerial
incentives. Its main component is the positive interaction between managerial incentives
and carbon pricing, which matters for two reasons: First, it drives the positive role of
managerial incentives, our main prediction. Second, this interaction means that the effect
of carbon pricing is expected to be augmented by green bonds.18
Appendix B assumes a more general functional form of the ∆c(i) function—including
the affine case of Corollary 1—that allows the explicit characterization of the equilibrium.
Appendix C extends our model to the presence of concerned investors.
In the next section, we will exploit variations in managerial incentives across sectors
and over time as well as changes in effective carbon prices across countries and over time
in order to validate the role of managerial incentives in expression (12).
18. The last component involves only managerial incentives. Although the main, interaction term is
proportional to the carbon price, this last term cannot be immediately interpreted as the effect that
managerial incentives would have if there were no carbon price. Indeed, as explained earlier, if the
carbon price was zero and there was no beneficial green projects (∆c(0) = a0), there would not be
any green projects at all, even with green bonds. Green finance might only play a role by itself, i.e.,
independently of carbon pricing, if there existed beneficial green projects (∆c(0) = a < 0).
IV. Empirical analysis of the role of managerial incentives and
carbon prices
In this section, we test our model’s main prediction: the role of managerial stock-price
sensitivity’s interaction with carbon prices in green bond issuance, beyond the usual role
of carbon pricing alone.
IV.A. Data
Our main sample is a panel that relates yearly data on individual firms, the green
bonds that they issued, and effective carbon prices that prevail in their countries. The
period under review is 2007-2019.
To assemble this sample, we start from green bonds data extracted from Bloomberg on
all corporate green bonds issued between January 2007 and December 2019: their issuer,
amount, yield, maturity, announcement and issuance dates. We use bonds’ information
from the non-profit Climate Bond Initiative (CBI) to eliminate non-certified bonds.
We consider all public firms in countries where green bonds have been issued and use
the above data to relate them to the volume of certified green bonds they issued every
year. Only a small number of firms have issued green bonds. Therefore, the data feature
not only differences in green bonds’ volume across firms that issue green bonds, but also
differences between firms that issue green bonds and those that do not.
Each firm is associated with the effective carbon price that prevails in its country.19
Effective carbon penalties consist not only of tradeable emission permit prices and car-
bon taxes, but also of all carbon-based fuel excise taxes. Estimates of effective carbon
prices and their coverage of various sources of CO2 emissions (road transport; non-road
transport; industry; agriculture; residential and commercial; and electricity) are provided
by the OECD for 2012, 2015, and 2018—“the most detailed and most comprehensive ac-
count of [the largest and most developed economies’] price of carbon emissions” (OECD,
19. We will address the case of multinationals that may be impacted by various carbon prices both
with firm fixed effects and with the addition of firms’ measures of foreign activities—see Table VII of
Appendix D.
2018). First, we use these estimates to derive the weighted average effective carbon price
in each country in 2012, 2015, and 2018, in current US . Second, we linearly interpolate
to obtain data for intermediate years.
Moreover, each firm is associated with annual financial data extracted from CRPS
and Compustat: market capitalization, book value, R&D expenditures, net debt issuance,
monthly traded number of shares, number of shares outstanding, as well as the percentage
of foreign sales. We add firms’ environmental scores and CO2 emissions from ASSET4.
Since our model predicts the proportion of green bonds in debt, we take the ratio
of firms’ volume of green bonds over their net debt issuance. For simplicity, we call the
obtained, standardized variable “Green bonds.”
Each firm is associated with its industry categories according to the Global Indus-
try Classification Standard (GICS). Firms are classified into the 69 industries that are
described in Appendix D.
Last but not least, the managerial sensitivity to firms’ stock price (“Incentives”) will
be captured by two complementary proxies. The first one is the wealth-performance
sensitivity, suggested by Edmans et al. (2009), and provided by Alex Edmans on a
yearly basis: the change in CEO wealth, following a 100 percentage point change in
firm value, scaled by annual flow compensation.20 This directly measures the weight of
stock components in managers’ financial compensation—an immediate interpretation of
the sensitivity of managers’ objective to the stock price of our model—in a way that
allows comparisons across firms and industries. Since our model captures managerial
incentives by the weight α(0,1), we divide the wealth-performance sensitivity by
its highest value in the sample. For simplicity, we call the obtained variable, “WPS.”
It varies significantly across industries, as Figure IV illustrates—see, moreover, the data
description of Appendix D. However, WPS is not available for firms based outside the US.
Therefore, we aggregate WPS at the industry and year level in the US, and extrapolate
20. CEO wealth includes shares and stock options, while compensation flows represent salary, bonuses,
and new grants of equity (Edmans et al., 2009).
it to the same industries in other countries. For example, we consider that the WPS
measure for the Automobiles industry in a given year is the same in the US and in
Our second, complementary proxy is based on stock share turnover, which is available
for all industries and countries. It reflects not only the focus of stock markets on short-
term executive results, but also the intensity of speculation and the risk of takeover. For
example, Summers and Summers (1989) suggest that stock share turnover is linked to
executive myopia. Cremers et al. (2020) confirm that it is associated with the presence of
short-term investors. To construct our share turnover variable, first, we divide, for each
firm and year, the average number of monthly traded shares by the number of shares
outstanding, and, second, we divide it by its highest value. We call it “Turnover.” Like
WPS, it varies significantly across industries—see Appendix D.
In Appendix D, we provide more practical details on the way we collected, and as-
sembled, green bond data, as well as summary statistics.
IV.B. Green bonds, managerial incentives, and carbon prices
We now test the main prediction of our theory, that is the positive role of managerial
incentives. More precisely, we seek to verify that, on average, in an industry, firms’
proportion of green bonds increases with managers’ sensitivity to their firms’ stock price,
in a way that is more pronounced as their country’s effective carbon price is higher.
For example, Figure IV shows the unconditional relationship between the Green bonds
and WPS variables in sectors that issue green bonds. It illustrates that sectors in which
managers’ pay is the most stock-price sensitive issue more green bonds.
Closer to our prediction (12), we estimate the following model in which the role of
Capital GoodsCapital GoodsCapital Goods
0 .005 .01
Green bonds
0 .005 .01 .015
Incentives (WPS)
Figure IV: Green bonds issuance and stock price sensitivity of managers’ compensation
managerial incentives includes the interaction with carbon prices:21
Green bondsi,t =β0+β1Carbon pricec(i),t1×Incentivesj(i),t1+β2Incentivesj(i),t1
+β3Controlsi,t1+Fixed effects +i,t.(13)
In this empirical model, the dependent variable is the proportion of green bonds issued
in year tby firm i. The independent variables of interest are the degree of managerial
incentives in industry j(i) and the effective carbon price that prevails in country c(i), both
taken at year t1. For managerial incentives, we will alternately use our two proxies.
Control variables include firm i’s market capitalization and its book-to-market ratio in
year t1 as well as its environmental score at year t1. Moreover, we include firm fixed
effects (FE), as well as industry, time-varying industry, and time-varying country fixed
We estimate the model of equation (E.1) using the method of least-squares with
21. We omit the autonomous role of carbon prices encompassed in i0in (12), because it is fully absorbed
by the time-varying country fixed effects.
standard errors clustered at the country level.22 Table I reports the estimation results.
It shows similar estimates whether Incentives =WPS or Incentives =Turnover. With
WPS, for example, we find coefficients for the terms that involve managerial incentives in
column (1) that are significantly different from zero at the 1% level and that have signs
in line with their theoretical counterparts in our model’s prediction (12).
Table I: Green bond issuance, managerial incentives, and carbon price
This table presents estimates from panel regressions of the proportion of green bonds issued by firms on
their countries’ effective carbon prices and proxies of managers’ stock-price sensitivity in their industries.
We control for firm size, book-to-market ratio, and environmental score. Standard errors are clustered
at the country level. *, **, and *** denote significance at the 10%, 5%, and 1%, respectively.
Green bonds
(1) (2) (3) (4)
Incentives (WPS) -0.505***
Carbon price ×Incentives (WPS) 0.017*** 0.024*
(0.006) (0.014)
Incentives (Turnover) -0.158***
Carbon price ×Incentives (Turnover) 0.006** 0.005***
(0.002) (0.002)
Controls Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes
Industry FE Yes No Yes No
Industry-year FE No Yes No Yes
Country-year FE Yes Yes Yes Yes
Observations 15011 15008 15148 15145
R20.335 0.358 0.335 0.358
22. In Appendix D, Table V shows estimation results when standard errors are clustered at the country
and industry level. Moreover, Table VI shows that our results carry over when the Incentives variables
are treated exactly like the αparameter of our model, rather than redefined as ˜αα/(2 α) as per
Corollary 1.
IV.C. The total role of managerial incentives
With estimated coefficients of Table I, the total role of Incentives variables appears
positive on average, that is, assuming the average effective carbon price in our sample
of 32; effective carbon prices may seem large because they include various excise taxes.
Yet, it is not statistically different from zero at this level.
In line with our formula (12), the results of Table I indicate that the total contribution
of managerial incentives in stronger as the carbon price is higher. In Appendix D, we
consider, for example, the average effective carbon price in the EU, where the green bond
market is the most developed, accounting for about 50% of the global current volume.
We find that the total role of Incentives is positive and significantly different from zero
at the 5 or 10% level respectively when Incentives=Turnover and Incentives=WPS.
For example, the Automobiles industry in Germany has issued green bonds in the
past few years. Firms in this sector have, on average, a middle-of-the-road stock-price
sensitivity: WPSAutomobiles = 0.0002. The effective carbon price of Germany in 2018 was
Carbon priceGermany,2018 = $74. Our estimates predict that firms in this industry issue
around 4% more green bonds than the average firm.23
In general, where carbon emissions are sufficiently penalized, firms issue more green
bonds in industries in which managers are more sensitive to their stock price. Moreover,
this effect is more pronounced as carbon effective prices become higher.
Our results stress the key role of the interaction between managerial incentives and
carbon prices, with important implications. First, it means that one should expect more
green bond issuance for a given carbon price, than there would be otherwise in absence
of certification. But, second, this interaction implies that green bonds’ effectiveness rely
on the level of carbon pricing.
The amplification effect is confirmed when industry-year fixed effects replace the In-
centives explanatory variables in columns (2) and (4) to capture other potentially omitted
23. (0.017×740.505)×0.0002
0.0035 '4.3%, where 0.0035 is the average proportion of green bond issuance across
firms in our sample, including many firms that have issued no green bond.
IV.D. Extensions
Firm level variation.—The model of Section III as well as the estimated equation
(E.1) assume that firms’ managers in the same industry and country are equally sensitive
to their firms’ stock price. Indeed, both WPS and Turnover vary significantly across
(US, for WPS) industries, and our main approach exploits this variation. However, firm
level variations in stock share turnover can also be exploited because it is available for
firms in all industries and countries. In Appendix D, we verify that the central role of
the interaction between carbon prices and managerial incentives survives when Turnover
is measured at the firm level, along with industry-year fixed effects—see Table VIII.
Multinationals.—Moreover, we have assumed that firms are only applied the carbon
price of the country where they are based. Although it seems questionable for multi-
nationals, one can defend our approach by invoking Ben-David, Kleimeier, and Viehs’
(2020) finding that it is environmental policies where multinationals are based that play
the most significant role. Besides firms’ characteristics captured by fixed effects in equa-
tion (E.1), Table VIII of Appendix D also verifies that similar results hold once firms’
shares of foreign sales, assets, and income, as measures of foreign activity, are included
in Controls.
V. Empirical analysis of green bonds’ effectiveness and stock
market announcement reactions
The theory presented in Section III deals with the reasons why firms issue green
bonds. Our model predicts the crucial role of managerial incentives and its interaction
with carbon prices, that we examine empirically in Section IV.
Our model further suggests that two main aspects underly the joint role of managerial
incentives and carbon prices. The first one is that green bonds effectively finance CO2
reduction. According to our theory, this is why carbon prices induce more green projects
and positive stock market reactions. The second one is that green bond issuance generates
positive stock returns. In our model, this is why managers issue these bonds.
In this section, we examine this two aspects.
V.A. Stock price reaction to green bonds
Positive stock price reaction to events like green bond announcements reflects that
these events reveal information about firms’ future profitability that investors could not
anticipate. In practice, these reactions are measured by abnormal stock returns around
green bond announcements, that are left unexplained by other factors.
We first express our model’s stock market reaction in terms of stock returns. Then,
we run an event-study estimation of abnormal returns.
V.A.1. Stock market returns at green bond issuance
In our model, firms’ date-0 stock returns are:
AkSk− S
S, k =G, B , (14)
depending on whether they choose a green or regular technology.
An immediate consequence of (9) is that the stock price of firms issuing green bonds
increases at a rate24
G= (1 ie)Se
which is the theoretical counterpart of abnormal stock returns at issuance.
24. In our model, conventional bonds signal that the project’s technology is not green, from which
investors infer that the project is relatively less efficient than in their ex ante assessment. Consequently,
the stock price of firms issuing conventional bonds decreases. (16) and (9) imply:
Our setting implies that abnormal returns (15) are negative. This effect, however, is proportional to the
proportion of green bonds, which is still small. In practice, regular bond news typically do not generate
noticeable abnormal returns (Antweiler and Frank, 2006). It is unrealistic to expect that conventional
bonds generate manifest negative stock market reactions unless green bonds become a prominent, almost
systematic, financing for green projects and carbon pricing become significantly more penalizing.
Therefore, positive stock returns at the issuance of green bonds AGonly differ across
firms issuing green bonds because the ex ante stock price Sincreases with V, the firm-
specific profit from regular activities, diluting stock returns induced by the project.25 In
other words, our model predicts that green bond announcement stock returns are higher
as green bond volumes are larger with respect to firms’ capitalization, which we now
attempt to verify.
V.A.2. Event-study estimation of abnormal stock returns at green bond an-
Various empirical studies examine abnormal stock returns at green bond issuance.
For example, Tang and Zang (2018), Baulkaran (2019), and Flammer (2021) find that
announcement abnormal returns are significant—see the introduction for more details.
Their results support the signaling mechanism at the core of our model.
Here, we replicate these results by estimating cumulative abnormal returns from an
event study analysis based on the global three-factor model of Fama and French (2012), as
we describe more precisely shortly below. Moreover, we compare the obtained abnormal
announcement returns according to the relative size of green bonds, featuring a positive
relationship between the former and the latter. This is in line with expressions (14)
and (16) of the stock market reaction to green bonds, also suggesting that the revealed
information has something to do with the size of financed projects.
Data.—We extend the panel data sample that we describe in Section IV to non certi-
fied green bonds. Moreover, we include daily stock prices to firms’ characteristics, from
which we derive stock market daily returns that we will denote by R. Our event study
will also use market factors data from Kenneth French.26
Event study analysis—Following prior studies (e.g., Tang and Zhang, 2020, and Flam-
mer, 2021), we explore how stock market reacts to green bond announcements using an
25. We omit the cost of certification, expertise, and monitoring, because it is probably negligible for
the large firms that currently issue green bonds. Admittedly, these costs may limit the supply of green
bonds by smaller firms.
26. See
event window around the announcement date. More specifically, our event window is
[-5;+5] with respect to the announcement date 0. For each company we estimate the
global three-factor model (Fama and French, 2012) to compute abnormal returns using
an estimation window of 250 days. We have a gap of 50 days between the end of the
estimation window and the beginning of the event window, so that the estimation window
can be represented by [-305;-55].
We estimate the following model:
Ri,t =β0,i +β1,iRm+β2,i Rs+β3,iRv+i,t ,(17)
where Ri,t is the daily stock market return of firm iat time tand Rm,Rs, and Rvare
the global market factor, size factor, and value factor respectively. We then obtain the
abnormal return as the difference between the observed daily stock return of firm iand
the estimated return:
ARi,t =Ri,t ˆ
Ri,t =ˆ
β0,i +ˆ
with the coefficients estimated from equation (17). Finally, the abnormal returns are
summed over the event window [-5;+5] to obtain the cumulative abnormal returns. This
is, for example, the method by which we estimate the stock returns that the Unilever
2014 green bond generated in our introductory example.
We now estimate similar returns for various categories of green bonds and firms that
issue them, in a same manner as, for example, Tang and Zang (2020) and Flammer
(2021). We consider a “large” category, that refers to certified green bonds whose dollar
amount divided by the issuing firm’s market capitalization is above the median. Table II
shows our results.27
27. The results presented in Table II are robust to an alternative event window specification (e.g.,
Table II: Event Study: Stock Market Reaction to Green Bonds
This table presents results on event studies around the date of green bonds’ announcement. The “fi-
nancial” group denotes companies with one of the following Bloomberg industry code: bank, financial,
REIT (real estate investment trust), or insurance. The “large” category refers to certified green bonds
of a dollar amount standardized by the firm’s market capitalization that is above the median.
All 0.68%∗∗∗ 432
Financials 0.65%181
Corporate 0.68%∗∗ 238
First Issue 0.75%∗∗ 215
Certified 0.75%∗∗∗ 282
Non-certified 0.46% 189
Large 0.75%∗∗ 141
On average, firms’ valuation increases by approximately 0.68% around the announce-
ment of green bond issuance. The effect is statistically significant at the 1% confidence
level and is concentrated among corporate firms’ issuance, first issuance, that is the first
time a firm issues a green bond, as well as large green bond issuance. As far as certified
green bonds are concerned, their effect is stronger in economic and statistical signif-
icance. Firms’ valuation increases by approximately 0.75% around announcements of
certified green bonds issuance. The effect is more pronounced for certified first-issuance
(almost 1% increase in firms’ value). Firms’ announcements of large green bonds have a
similar effect, in line with our model.
V.B. CO2 emissions after green bonds, and green bond public policies
We now examine the relationship between green bond issuance and subsequent CO2
emissions, exploiting the implementation of public policy supports to green bonds in some
In a recent yet influential paper, Flammer (2021) uses a matching method to show
that firms issuing certified green bonds reduce their CO2 emissions by 13% over the course
of the next two years. This result rules out the greenwashing hypothesis. The following
analysis seeks to further suggest a direction of causality, consolidating Flammer’s result.
Our identification strategy relies on the implementation of green bond policies in
Japan, Singapore, Hong-Kong, and Malaysia. These policies amount to a public support
to the use of green bonds, irrespective of firms’ current and future CO2 emissions or
of other factors of firms’ environmental performance. Firm-level data for companies in
Japan, Singapore, Hong-Kong and Malaysia are not sufficient so we would not be able to
systematically control for firms’ characteristics that might affect green bond issuance. We,
therefore, use a country-level aggregated version of the sample presented in Subsection
Data.—There are two basic differences between the firm-level sample presented in
Subsection IV.A and the sample that we use here. First, we construct the latter by ag-
gregating volumes of green bonds at the country level, which we divide by firms’ bond net
issuance in each country. Second, for each country and year, we indicate whether public
policies supporting the use of green bonds are implemented. From 2017 onwards, Japan,
Singapore, Hong-Kong and Malaysia implemented programs to support the issuance of
green bonds, helping issuers face various difficulties related to the certification process,
disclosure requirements, and monitoring of the financed projects—see the Climate Bonds
Initiative’s (2018) report. In the baseline model of Section III, the cost of certified green
projects may be interpreted to include issuance, certification and monitoring costs, that
are alleviated by green bond supporting policies. These policies all amount to a form
of green bond subsidy. There is no available information on the intensity of these poli-
cies. Therefore, we include a dummy variable taking value one if and when a country
has a green bond policy, which we call “Policy.” This variable is an indicator that takes
the value 1 for China post-2016, Malaysia and Singapore post-2017 and Hong Kong and
Japan post-2018. This variable will be used as an instrument.
Instrumental variable analysis.—In the first stage, we show how green bonds were
affected by green bond policy supports. Formally, we estimate the following model:
Green bondsc,t =β0+β1Policyc,t1+β2Controlsc,t1+Fixed effects +c,t.(20)
In this model, the index crefers to countries. The dependent variable is the proportion
of green bonds issued in year tby firms based in country c. The independent variable of
interest is the Policy variable evaluated in country cand year t1. Moreover, controls
consist of Gross Domestic Product (GDP), GDP per capita, total debt, and carbon price
in country cand year t1.
We estimate all coefficients by the method of least squares with standard errors clus-
tered at the country level. The results are presented in column (1) of Table III. They
show that the coefficient β1is positive and significantly different from zero at the 5%
level. The implementation of green bond policy supports was associated with an increase
in green bonds in the next year.
The underidentication test strongly rejects the null of no correlation between our
Policy instrument and Green bonds. The first-stage F-statistic is 123.01, significantly
higher than Stock and Yogo’s (2005) critical value of 16.38 for a 10% maximal bias of the
instrumental variable estimator relative to the bias of ordinary least squares.
Table III: Green bond policy and country-level CO2 emissions
This table presents estimates from panel regressions of countries’ CO2 emissions on the proportion of
green bonds issued by countries, instrumented by whether policies supporting green bonds have been
implemented. We control for GDP, GDP per capita, the total volume of bonds issued by each country in
each year, as well as the carbon price. Standard errors are clustered by country. *, **, and *** denote
significance at the 10%, 5%, and 1%, respectively.
1st stage 2nd stage
Green bonds CO2
(1) (2) (3)
Policy (1 year), instrument 0.120**
Instrumented Green bonds, (1 year) -0.606** -0.623**
(0.272) (0.260)
Controls Yes Yes Yes
Year FE Yes No Yes
Observations 211 211 211
R20.703 0.701
We now take a two-stage-least-squares estimation approach to show how the instru-
mented values of Green bonds for country cand year t, denoted by \
Green bonds are linked
to CO2 emissions. The model that we now estimate is:
CO2c,t+1 =β0+β1\
Green bondsc,t +β2Controlsc,t +Fixed effects +c,t.(21)
Our dependent variable is total CO2 emissions of all firms in country cin year t. Besides
the main independent variable of interest \
Green bonds for country cand year t1, we
include GDP, GDP per capita, total debt, and carbon price as controls and year fixed
effects as in the first stage. We cluster standard errors at the country level.
The results are presented in columns (2)-(3) of Table III, showing that the coefficient of
the predicted green bonds is negative and significantly different from zero at the 5% level.
Our estimates mean that firms on average reduce their CO2 emissions by approximately
15% following an increase in green bonds issuance of 1 percentage point.
VI. Conclusion
This paper examines why firms issue green bonds. From the formal analysis of Section
III, and empirical results of Sections IV and V, we draw the following conclusions. First,
although voluntary, certified green bonds can induce firms to commit to effective CO2
reducing projects. This is because firms’ announcement of certified green projects convey
positive information about the profitability of these projects, as abnormal stock returns
reflect. Second, perhaps surprisingly, firms’ incentives to issue green bonds is likely a
matter of financial interest and short-termism. Third, green bonds are complementary to
carbon pricing, with important practical implications. First and foremost, green bonds do
not help governments avoid carbon penalties; on the contrary, the latter are instrumental
in the effectiveness of the latter. At the same time, if carbon prices are sufficiently high,
green bonds are likely to make them more effective.
Our theory does not rely on investors’ concern for environmental performance, but
easily accommodates it. Although investors’ concern has hitherto not played an apparent
role, some expect it to change in the medium run.28 Appendix C shows how our model
carries over to the case of concerned investors on both bond and stock markets, generating
respectively a green bond yield spread and an augmented abnormal stock returns at green
bond issuance that would induce additional incentives to undertake green projects. It
shows that the presence of concerned investors would not qualitatively change our baseline
28. See, for example, Financial Times, January 4, 2021, “Analysts expect as much as 500bn of green
bonds in bumper 2021.”
model’s prediction as to the role of managerial incentives.
To our knowledge, there exists no study on the economic mechanisms that drive green
bond issuance, neither in the finance literature, nor in the environmental economics and
policy literature, despite the fact that green bonds are becoming increasingly prominent
and the fact that financial institutions are currently paying a lot of attention to green
finance instruments. This paper’s attempt to provide consistent theoretical and empirical
foundations to sustainable green finance instruments may be useful not only to economic
applications to new green finance markets, but also to the understanding and empirical
assessments of the role that they may play in the structure of climate policy in contexts
in which carbon pricing is a restricted option.
If carbon pricing remains a limited option for governments, the urgency of the climate
challenge will require that all alternatives to mitigate CO2 emissions be examined. Our
analysis points to the critical, perhaps paradoxical, role of managerial incentives, and
calls for more research on the effective design of managerial compensation as a climate
policy regulation.
Antweiler, W., and M. Frank (2006), “Do US Stock Markets Typically Overreact to
Corporate News Stories?,” SSRN Working Paper 878091.
Baker, M., D. Bergstresser, G. Serafeim, and J. Wurgler (2020), “Financing the Response
to Climate Change: The Pricing and Ownership of U.S. Green Bonds,” Harvard Business
School mimeo.
Barrage, L., and J. Furst (2019), “Housing Investment, Sea Level Rise, and Climate
Change Beliefs,” Economics Letters, 177: 105-108.
Baulkaran, V. (2019), “Stock Market Reaction to Green Bond Issuance,” Journal of Asset
Management, 20: 331-340.
Bebchuk, L.A. (2021), “Don’t Let the Short-Termism Bogeyman Scare You,” Harvard
Business Review, January-February.
Ben-David, I., S. Kleimeier, and M. Viehs (2020), “Exporting Pollution,” NBER Working
Paper 25063.
enabou, R., and G. Laroque (1992), “Using Privileged Information to Manipulate Mar-
kets: Insiders, Gurus, and Credibility,” Quarterly Journal of Economics, 107: 921-958.
enabou, R., and J. Tirole (2010), “Individual and Corporate Social Responsibility,”
Economica, 77: 1-19.
Bolton, P., J. Scheinkman, and W. Xiong (2006), “Executive Compensation and Short-
Termist Behaviour in Speculative Markets,” Review of Economic Studies, 73: 577-610.
Bonneton, N. (2020), “The Market for Ethical Goods,” Toulouse School of Economics
Bonroy, O., and C. Constantatos (2015), “On the Economics of Labels: How Their
Introduction Affects the Functioning of Markets and the Welfare of All Participants,”
American Journal of Agricultural Economics, 97: 239-259.
Bouvard, M., and R. Levy (2018), “Two-Sided Reputation in Certification Markets,”
Management Science, 64: 4755-4774.
Casamatta, C., and S. Pouget (2015), “Fund Managers’ Contracts and Short-Termism,”
Toulouse School of Economics mimeo.
Central Banking (2021), “ECB Steps Up Work on Climate Change,” available at https:
Chava, S. (2014), “Environmental Externalities and Cost of Capital,” Management Sci-
ence, 60: 2223-2247.
Chiang, J. (2017), “Growing the US Green Bond Market,” California State Treasurer’s
Office, Sacramento, CA.
Climate Bonds Initiative (2018), Creating green Bond Market – Insights, Innovation, and
Tools from Emerging Markets.
Cremers, M., A. Pareek, and Z. Sautner (2020), “Short-Term Investors, Long-Term In-
vestments, and Firm Value: Evidence from Russell 2000 Index Inclusions,” Management
Science, 66: 4525-4551.
Daubanes, J., and J.-C. Rochet (2019), “The Rise of NGO Activism,” American Eco-
nomic Journal: Economic Policy, 11: 183-212.
Denicol`o, V. (2008), “A Signaling Model of Environmental Overcompliance,” Journal of
Economic Behavior and Organization, 68: 293-303.
Diamond, P. (2006), “Optimal Tax Treatment of Private Contributions for Public Goods
with and without Warm Glow Preferences,” Journal of Public Economics, 90: 897-919.
Dimanchev, E.G., and C.R. Knittel (2020), “Trade-offs in Climate Policy: Combining
Low-Carbon Standards with Modest Carbon Pricing,” MIT CEEPR Working Paper 2020-
Eckbo, B. (1986), “Valuation Effects of Corporate Debt Offerings,” Journal of Financial
Economics, 15: 119-151.
Edmans, A., V.W. Fang, and K.A. Lewellen (2017), “Equity Vesting and Investment,”
Review of Financial Studies, 30: 2229-2271.
Edmans, A., X. Gabaix, and A. Landier (2009), “A Multiplicative Model of Optimal
CEO Incentives in Market Equilibrium,” Review of Financial Studies, 22: 4881-4917.
Edmans, A. (2020), Grow the Pie: How Great Companies Deliver Both Purpose and
Profit, Cambridge University Press.
Environmental Finance (2017), “Of Repsol and Reputation,” available at https://www.
Farhi, E., J. Lerner, and J. Tirole (2005), “Certifying New Technologies,” Journal of the
European Economic Association, 3: 734-744.
Fama, E.F., and K.R. French (2012), “Size, Value, and Momentum in International Stock
Returns,” Journal of Financial Economics 105: 457-472.
Flammer, C. (2021), “Corporate Green Bonds,” Journal of Financial Economics, 142:
Gabaix, X., P. Gopikrishnan, V. Plerou, and H.E. Stanley (2006), “Institutional Investors
and Stock Market Volatility,” Quarterly Journal of Economics, 121: 461-504.
Gabaix, X., and A. Landier (2008), “Why Has CEO Pay Increased So Much?,” Quarterly
Journal of Economics, 123: 49-100.
Gibson, R., P. Krueger, and S.F. Mitali (2020), “The Sustainability Footprint of Insti-
tutional Investors: ESG Driven Price Pressure and Performance,” European Corporate
Governance Institute – Finance Working Paper 571.
Goergen, M., and L. Renneboog (2011), “Managerial Compensation,” Journal of Corpo-
rate Finance, 17: 1068-1077.
Gollier, C., and S. Pouget (2009), “Shareholder Activism and Socially Responsible In-
vestors: Equilibrium Changes in Asset Prices and Corporate Behavior,” TSE Working
Paper 09-081.
Gopalan, R., T. Milbourn, F. Song, and A.V. Thakor (2014), “Duration of Executive
Compensation,” Journal of Finance, 69: 2777-2817.
Harrison C., C. Partridge, and A. Tripathy (2020), “What’s in a Greenium: An Analysis
of Pricing Methodologies and Discourse in the Green Bond Market,” Journal of Environ-
mental Investing, 10: 64-76.
Hart, O., and L. Zingales (2017), “Companies Should Maximize Shareholder Welfare not
Market Value,” European Corporate Governance Institute – Finance Working Paper 521.
Hartzell, J.C., and L.T. Starks (2003), “Institutional Investors and Executive Compen-
sation.” Journal of Finance, 58: 2351-2374.
Heal, G. (2005), “Corporate Social Responsibility: An Economic and Financial Frame-
work,” Geneva Papers on Risk and Insurance, 30: 387-409.
Kapraun, J., and C. Scheins (2020), “(In)-Credibly Green: Which Bonds Trade at a
Green Bond Premium?,” Goethe University Frankfurt mimeo.
Khanna, M., and L.A. Damon (1999), “EPA’s Voluntary 33/50 Program: Impact on Toxic
Releases and Economic Performance of Firms,” Journal of Environmental Economics and
Management, 37: 1-25.
Klassen, R.D., and C.P. McLaughlin (1996), “The Impact of Environmental Management
on Firm Performance,” Management Science, 42: 1199-1214.
Kotchen, M.J., and C. Costello (2018), “Maximizing the Impact of Climate Finance:
Funding Projects or Pilot Projects?,” Journal of Environmental Economics and Manage-
ment, 92: 270-281.
Krueger, P., Z. Sautner, D.Y. Tang, and R. Zhong, (2021), “The Effects of Mandatory
ESG Disclosure Around the World,” Swiss Finance Institute Research Paper 21-44.
Jacobsen M.R., C.R. Knittel, J.M. Sallee, and A.A. van Benthem (2021), “Sufficient
Statistics for Imperfect Externality-Correcting Policies,” Journal of Political Economy,
Landier, A., and S. Lovo (2021), “Socially Responsible Finance: How to Optimize Im-
pact?,” HEC Paris mimeo.
Ladika, T., and Z. Sautner (2020), “Managerial Short-Termism and Investment: Evidence
from Accelerated Option Vesting,” Review of Finance, 24: 305-344.
Lerner, J., and J. Tirole (2006), “A Model of Forum Shopping,” American Economic
Review, 96: 1091-1113.
Lian, C., and Y. Ma (2021), “Anatomy of Corporate Borrowing Constraints,” Quarterly
Journal of Economics, 136: 229-291.
Lyon, T.P., and J.W. Maxwell (2003), “Self-Regulation, Taxation and Public Voluntary
Environmental Agreements,” Journal of Public Economics, 87: 1453-1486.
Magill, M., M. Quinzii, and J.C. Rochet (2015), “A Theory of the Stakeholder Corpora-
tion,” Econometrica, 83: 1685-1725.
Maxwell, J.W., T.P. Lyon, and S.C. Hackett (2000), “Self-Regulation and Social Welfare:
The Political Economy of Corporate Environmentalism,” Journal of Law and Economics,
43: 583-617.
Marinovic, I., and F. Varas (2019), “CEO Horizon, Optimal Pay Duration, and the
Escalation of Short-Termism,” Journal of Finance, 74: 2011-2053.
Mikkelson, W.H., and M.M. Partch (1986), “Valuation Effects of Security Offerings and
the Issuance Process,” Journal of Financial Economics, 15: 31-60.
OECD (2018), “Effective Carbon Rates 2018 – Pricing Carbon Emissions Through Taxes
and Emissions Trading,” available at
Pastor, L., R.F. Stambaugh, and L.A. Taylor (2021), “Dissecting Green Returns,” Becker
Friedman Institute Working Paper 2021-70.
van der Ploeg, R., and A. Rezai (2019), “Stranded Assets in the Transition to a Carbon-
Free Economy,” University of Oxford mimeo.
Polsky, G.D., and A.C.W. Lund (2013), “Can Executive Compensation Reform Cure
Short-Termism?,” Issues in Governance Studies, 58: 1-9.
Schleifer, A., and R.W. Vishny (1990), “Equilibrium Short Horizons of Investors and
Firms,” American Economic Review, 80: 148-153.
Sobel, J. (1985), “A Theory of Credibility,” Review of Economic Studies, 52: 557-573.
Stavins, R.N. (2008), “Addressing Climate Change with a Comprehensive US Cap-and-
Trade System,” Oxford Review of Economic Policy, 24: 298-321.
Stein, J.C. (1988), “Takeover Threats and Managerial Myopia,” Journal of Political Econ-
omy, 96: 61-80.
Stein, J.C. (1989), “Efficient Capital Markets, Inefficient Firms: A Model of Myopic
Corporate Behavior,” Quarterly Journal of Economics, 104: 655-669.
Stock, J.H., and M. Yogo (2005), “Testing for Weak Instruments in Linear IV Regression,”
in Andrews, D.W.K., and J.H. Stock (Eds.), Identification and Inference for Econometric
Models: Essays in Honor of Thomas J. Rothenberg: 80-108, Cambridge University Press.
Summers, L.H., and V.P. Summers (1989), “When Financial Markets Work Too Well: A
Cautious Case for a Securities Transaction Tax,” Journal of Financial Services Research,
3: 261-286.
Sustainalytics (2019), “Apple Green Bond Framework,” available at https://www.sustainalytics.
Tang, D.Y., and Y. Zhang (2020), “Do Shareholders Benefit from Green Bonds?,” Journal
of Corporate Finance, 61: 101427.
Thanassoulis, J. (2013), “Industry Structure, Executive Pay, and Short-Termism,” Man-
agement Science, 59: 402-419.
Tietenberg, T. (1998), “Disclosure Strategies for Pollution Control,” Environmental and
Resource Economics, 11: 587-602.
Zerbib, O.D. (2019), “The Effect of Pro-Environmental Preferences on Bond Prices: Ev-
idence from Green Bonds,” Journal of Banking and Finance, 98: 39-60.
Online Appendix to
“Why Do Firms Issue Green Bonds?”
Julien Xavier Daubanes
University of Geneva (GSEM), MIT (CEEPR), and CESifo
E-mail address:
Shema Fed´eric Mitali
Swiss Federal Institute of Technology in Lausanne (CDM), and Swiss Finance Institute
E-mail address:
Jean-Charles Rochet
University of Geneva (GSEM), Swiss Finance Institute,
and Toulouse School of Economics
E-mail address:
November, 2021
A. Baseline model: Proofs
Proof of Lemma 1
The manager of the firm with the project of type ichooses the green technology if and
only if UG(i)≥ UB(i) in which SGand SBare parametric. By the managerial objective
(2), this is equivalent to the condition (4) expressed in the lemma.
The rest of the Lemma follows directly from the implicit definition of iein (4).
Proof of Lemma 2
Firms opting for a conventional technology bear business-as-usual cost cB. Therefore,
their stock price is
1 + ρ0,(A.1)
which only varies by V, and is further independent of the bond market equilibrium.
As for firms that issue green bonds, investors infer that they have the most profitable
green projects, as per Lemma 1, even though they cannot perfectly assess their costs.
Consider a given proportion of green bonds ie. The resulting stock price of green firms
will be
1 + ρ,(A.2)
which decreases with iebecause an expansion of green projects means that green projects
are less efficient on average.
Moreover, in an interior equilibrium, all projects iiewill be green and their firms’
stock will be priced (A.2); otherwise, projects will be conventional and their firms’ stock
will be priced (A.1).
The stock market reaction ∆S=SG−SB. Using (A.2) and (A.1), we obtain expression
The expression of the ex ante stock price (8) is the weighted average of anticipated
stock prices for firms making a green projects and others. It holds because all projects
are undertaken. (7) follows from (8) and (9). Let us now show that (8) is maximum when
the anticipated proportion of green projects ieis the proportion of green projects that
would prevail in absence of green bonds i0. In (8), only the term on the right depends
on ie. Its derivative is equal to τxc(ie). It is equal to zero for ie=i0. When ieis
anticipated to be greater, as will turn our to be true in equilibrium, this derivative will
be negative.
Note that if investors anticipate the corner equilibrium ie= 1, then, by (8), the ex
ante stock price becomes S=SG, so that they will be no stock market reaction to green
Proof of Proposition 1
(4) and (9) are relationships between the proportion of green bonds ieand the stock
market announcement reaction ∆Sethat are respectively increasing and decreasing.
Therefore, their intersection yields a unique proportion of green bonds
ie=ie(α, τ ),
and expression (10) of the equilibrium stock market reaction to green bond announcement.
Since ∆c(ie)>E[∆c(i)|iie], (10) implies that this reaction is always positive.
Moreover, (4) and (9) implies that ie(α, τ ) can be characterized by (11). Since
both ∆c(ie) and E[∆c(i)|iie] are increasing in ieand ∆c(ie)>E[∆c(i)|iie], the
comparative-static effects of αand τfollows, as illustrated in Figure III.
The main prediction of our theory is the positive total effect of managerial incentives
on green bond issuance. Total differentiating (11) with respect to ieand α, holding τ
unchanged, and rearranging terms, we obtain:
∂ie(α, τ )
∂α =c(ie)E[∆c(i)|iie]
(1 α)∆c0(ie) + αdE[∆c(i)|iie]
which, for example, depends on carbon pricing positively through the stock market reac-
tion of the numerator.
Similarly, one obtains:
∂ie(α, τ )
∂τ =x
(1 α)∆c0(ie) + αdE[∆c(i)|iie]
establishing that carbon pricing contributes to increase the proportion of green projects,
in a way that depends on managerial incentives.
Conditions for the equilibrium to be interior, 0 < ie<1, are i0>0 and ie<1.
The Former is equivalent to τx > c(0); this is, the carbon price is sufficiently high
to induce some green projects even in the absence of green bond certification and stock
market reaction. The former, ie<1 requires first that some projects are not certified
green, i.e., E[πB] = YRcBτxB0, and, second, that ieas characterized by (11)
is less than 1, which is guaranteed by E[∆c(i)|i1] = R1
0c(i)di > τx.
The linear expression (12) is obtained under the affine assumption that ∆c(i) = a+bi,
b > 0, is derived in the more general case solved explicitly in Appendix B.
B. Baseline model: An example with explicit solutions
In this appendix, we assume the following functional form for the additional cost of
green projects:
c(i)a+biγ, b, γ > 0.
The resulted expected cost of certified green projects iieis:
E[∆c(i)|iie] = a+b(ie)γ
γ+ 1.
The supply of green bonds (4) becomes:
(1 + ρ)∆S
The stock market reaction to green bonds (9) becomes:
(1 + ρ)∆Se=bi0γ(ie)γ
γ+ 1.(B.2)
Replacing (B.2) into (4) and rearranging, one obtains the rational expectation equi-
librium green bond proportion:
ie=γ+ 1
γ+ 1 αγ 1
in which the amplification effect of green bonds is reflected by the fact that the term
between brackets is greater than unity. Moreover, the equilibrium proportion of green
bonds is increasing in αin a way that is more pronounced when τis higher.
Finally, when γ= 1, as in Corollary 1,
from which one can directly recover the linear prediction (12).
C. Extension to investors’ concern
In this appendix, we extend the model of Section III to the presence of bond and
stock investors that value the environmental impact of firms’ projects. We retain the
assumptions of Section III, apart from the following changes.
C.A. Green finance and firms’ problem
We assume that green and conventional bonds have different yields. Like in Section
III, we consider that conventional bonds repay RB=R1 + r, given by the exogenous
interest rate r. By contrast, we consider that the yield of green bonds is lower than r, by
a spread s0, so that
The additional profit generated at date t= 1 by an incremental project of type
i[0,1] with technology k=G, B becomes, instead of (1):
πk(i) = YRkck(i)τxk+εk(i).(C.1)
C.B. Green bond supply and demand
Managers’ maximization of (2) by choice of their project’s technology k=G, B now
takes the green bond spread as given. It follows that the marginal project that is certified
through a green bond is now characterized by, instead of (4):
(1 α) (∆c(ie)τxs) = α(1 + ρ)∆S.(C.2)
In reality, investors are certainly heterogenous as far their preference for environmen-
tal performance is concerned; it is the marginal investor that matters. However, the
simplifying assumption that all investors are similarly concerned is sufficient to figure out
the consequence of a green bond spread. We assume that bond investors attach the same
warm glow value θBx0 to the CO2 reduction due to certified green bonds so that,
in equilibrium, the no-arbitrage condition r=rGθBxholds, implying the following
demand condition for green bonds:
s=θBx. (C.3)
Following (C.1) and (C.3), the results of Lemma 1 should be amended as follows: The
equilibrium green bond spread is se=θBx; for a given stock market reaction ∆S, the
proportion of green bonds issued increases with the degree of bond investors’ concern θB
through the green bond yield spread, and it is larger than in the absence of green bonds
if α > 0 and ∆S>0 or if θB>0.
C.C. Stock market reaction to green bonds
Shareholders may be interested in the overall environmental performance of a firm, as
reflected, in reality, by its environmental rating. We assume, in a similar way as we do
for bond investors, that stock investors attach the same value θSx0 to firms’ CO2
reduction due to a certified green project. Equilibrium stock prices, therefore, become:
Sk=V+E[πk(i)|k] + θSx
1 + ρ, V 0, i [0,1], k =G, B. (C.4)
Lemma 2 should in this context be amended as follows: At date t= 0, the stock price
of firms issuing green bonds increases with respect to others by:
(1 + ρ)∆Se=τxE[∆c(i)|iie] + θSx; (C.5)
for a given proportion of green bonds, this stock market reaction is more pronounced
when the degree of stock investors’ concern θSbecomes higher.
An immediate consequence of (C.5) is that the stock price of firms issuing green bonds
increase at a rate:
G= (1 ie)Se
where Ak(Sk− S)/S,k=G, B.AGis the theoretical counterpart of abnormal stock
returns at issuance. (C.5) and (C.6) imply that the environmental rating effect of stock
investors’ concern on stock prices depends on the size of the financed project with respect
to issuing firms’ capitalization.
C.D. Equilibrium proportion of certified green projects
Finally, the intersection of relationships (C.2), with (C.3), and (C.5) determines a
unique rational expectation equilibrium proportion of green bonds
ie=ie(α, τ, θB, θS),
as depicted in Figure VI, which is the counterpart of Figure III in the presence of con-
cerned investors. The thick curves are those of Figure III (θB=θS= 0), and the dashed
curves result from the introduction of concern parameters θB, θS>0.
α(1 + ρ)∆S
α(1 + ρ)∆Se
(1 α) (∆c(i)x(τ+θB))
Figure V: Equilibrium proportion of green projects with concerned investors
The following results are obtained, which are the counterpart of Proposition 1 and
Corollary 1. They show how the equilibrium characterized in Section III and its linear
testable prediction accommodates the presence of concerned investors.
Proposition 2 (Green bonds equilibrium with concerned investors).
1. The equilibrium with rational expectations exists and is unique;
2. In this equilibrium:
(a) The stock market reaction to green bonds is Se>0;
(b) The resulting proportion of green bonds increases with the industry’s man-
agerial sensitivity to the stock price α, the carbon price τ, and the degrees
of investors’ concern θBand θS;
3. Assume that the green technology cost takes the affine form c(i) = a+bi, with
b > 0, then the closed-form linear expression follows:
| {z }
| {z }
managerial incentives
[(1 α)θB+αθS]
| {z }
investors’ concern
To sum up, the role of managerial incentives in our theory carries over to the presence
of concerned investors, although investors’ concern shifts the demand of, and the stock
market reaction to, green bonds, in a way that strengthens firms’ incentives to undertake
green projects certified through green bonds.
D. Empirics
Mode details on the green bond data
Green bonds’ data are extracted from the fixed income Bloomberg database. Cor-
porate green bonds are indicated by the use of proceeds “Green Bond/Loan” and asset
class “Corporate.” Identifiers used are bond ISIN, company ISIN, and CUSIP. The unique
bond ISIN identifier is used to merge Bloomberg green bonds’ data with CBI’s certifica-
tion information.
We include all public firms with codes 10 and 11, which makes a total for our analysis
of 19844 distinct firms. We define firms’ net debt following Lian and Ma (2021) as
the long-term debt issuance (Compustat item DLTIS) minus long-term debt reduction
(Compustat item DLTR).
We merge the bonds information with CRSP and Compustat using the firms’ CUSIP
and ISIN identifiers.
We define industries using the Global Industry Classification Standard (GICS), which
is provided with the green bond dataset. It is divided into 69 industries which can also
be categorized into 24 industry groups and 11 sectors.
Table IV below shows summary statistics of all green bonds obtained from the Bloomberg
database. Panel A shows a stable growth in the volume of green bonds issued from 2007
to 2019, which is also manifest in Figure 1 of the introduction. Most green bonds are
issued by private industrial companies followed closely by banks. China and the United
States lead the issuance of green bonds.
Other variables’ description
This section describes firm and country-level variables’ sources and constructions.
Book-to-market ratio.—To obtain book equity from Compustat, we subtract from the
shareholders’ equity the preferred stock value, using redemption, liquididating or carrying
value in that order (items PSTKRV, PSTKL, PSTKQ). For shareholders’ equity we use
the item SEQ, or Total Common Equity plus Preferred Stock Par Value (CEQ, PSTKQ)
if SEQ is missing and Total Assets minus Total Liabilities minus Minority Interest if CEQ
or PSTKQ is missing, using items ATQ, LTQ, and MIBQ. We then divide by the market
value of the firm, which is obtained as the number of shares oustanding multiplied by the
stock price, as in CRSP, and Compustat Global for international firms.
Carbon price.—It is a weighted average of effective carbon rates across all sectors
(Road transport; Non-road transport; Industrial facilities; Households, commercial and
public services; Electricity) weighted by the amount of emissions of each sector and each
coverage type (permit prices and/or taxes). We obtain these data from the OECD for
years 2012, 2015 and 2018 and we linearly interpolate the resulting estimates to obtain
data for intermediate years.
CO2 emissions.—We collect firm-level carbon emissions from ASSET4 (item EN-
ERDP023, Total Carbon dioxide (CO2) and CO2 equivalents emission in tonnes, i.e.,
scopes 1 and 2).
Environmental score.—We collect firm-level environmental scores from ASSET4 (item
ENSCORE; notes: ”Refinitiv’s Environment Pillar Score is the weighted average relative
rating of a company based on the reported environmental information and the resulting
three environmental category scores”).
Exchange rate.—We use yearly exchange rates from OECD to convert all carbon
prices to US denominated prices. More specifically, the exchange rate is the price of one
country’s national currency units in relation to US as of the end of each year.
Foreign sales, assets, an income.—We collect the proportion of foreign sales, assets,
and income from Worldscope, which is defined as international sales, assets, and income,
divided by net sales, total assets or revenues (item WC08731).
GDP.—We collect international data for the GDP from the World Bank.
Market capitalization.—We compute market capitalization as the number of shares
outstanding multiplied by the stock price using CRSP and Compustat Global data.
Net debt issuance.—We define net debt issuance following Lian and Ma (2021). We
compute it as long-term debt issuance (Compustat item DLTIS) minus long-term debt
reduction (Compustat item DLTR).
R&D expenditures.—Following prior studies (e.g., Ladika and Sautner, 2020; Cremers
et al., 2020), we collect R&D expenditures to capture long-term investments and thus
proxy for the opposite of short-termism. We collect R&D expenditures from Compustat
(item XRD).
Scaled WPS.—This measure is obtained from Alex Edmans’ website. It is defined
as the dollar change in CEO wealth for a 100 percentage point change in firm value,
standardized by the annual flow compensation (Edmans et al., 2009).
Share turnover.—For each firm and each year, it is the sum the monthly number
of shares traded in a given year (trading volume), divided by the number of shares
outstanding as of the end of the year. We take the average of this ratio for each industry
and year.
Trading volume.—We collect trading volume from CRSP and Compustat Global and
adjust it for stock splits.
Total debt (country-level).—We collect country-level total debt from the BIS database.
We consider gross issues of debt in a given country (domestic market) and year by all
institutions except governments, central banks, and international institutions.
E. Industry classification
Code Industry
code101010 Energy Equipment and Services
code101020 Oil, Gas and Consumable Fuels
code151010 Chemicals
code151020 Construction Materials
code151030 Containers and Packaging
code151040 Metals and Mining
code151050 Paper and Forest Products
code201010 Aerospace and Defense
code201020 Building Products
code201030 Construction and Engineering
code201040 Eletrical Equipment
code201050 Industrial Conglomerates
code201060 Machinery
code201070 Trading Companies and Distributors
code202010 Commercial Services and Supplies
code202020 Professional Services
code203010 Air Freight and Logistics
code203020 Airlines
code203030 Marine
code203040 Road and Rail
code203050 Transportation Infrastructure
code251010 Auto Components
code251020 Automobiles
code252010 Household Durables
code252020 Leisure Products
code252030 Textiles, Apparel and Luxury Goods
code253010 Hotels, Restaurants and Leisure
code253020 Diversified Consumer Services
code255010 Distributors
code255020 Internet and Direct Marketing Retail
code255030 Multiline Retail
code255040 Specialty Retail
Code Industry
code301010 Food and Staples Retailing
code302010 Beverages
code302020 Food Products
code302030 Tobacco
code303010 Household Products
code303020 Personal Products
code351010 Health Care Equipment and Supplies
code351020 Health Care Providers and Services
code351030 Health Care Technology
code352010 Biotechnology
code352020 Pharmaceuticals
code352030 Life Sciences Tools and Services
code401010 Banks
code401020 Thrifts and Mortgage Finance
code402010 Diversified Financial Services
code402020 Consumer Finance
code402030 Capital Markets
code402040 Mortgage Real Estate Investment Trusts (REITs)
code403010 Insurance
code451020 IT Services
code451030 Software
code452010 Communications Equipment
code452020 Technology Hardware, Storage and Peripherals
code452030 Electronic Equipment, Instruments and Components
code453010 Semiconductors and Semiconductor Equipment
code501010 Diversified Telecommunication Services
code501020 Wireless Telecommunication Services
code502010 Media
code502020 Entertainment
code502030 Interactive Media and Services
code551010 Electric Utilities
code551020 Gas Utilities
code551030 Multi-Utilities
code551040 Water Utilities
code551050 Independent Power and Renewable Electricity Producers
The following table presents some summary statistics, including Edmans’ original
wealth-price sensitivity and share stock turnover, before they are normalized to fit with
our model.
Table IV: Summary Statistics
This table presents summary statistics from Bloomberg’s Corporate Green Bonds data. The sample is
restricted to bonds with a green bond indicator that equals one according to Bloomberg and with a use
of proceeds that must include Green Bond/Loan.
Panel A: Distribution of Green Bonds by Issuance Year
N Total (MM)
2007 1 808
2008 7 427
2009 13 920
2010 50 4,229
2011 22 975
2012 21 2,047
2013 36 13,642
2014 123 31,314
2015 301 43,758
2016 225 85,477
2017 377 103,996
2018 457 113,946
2019 562 167,189
Panel B: Distribution of Green Bonds by Company Type
N Total (MM)
N/A 177 24,068
Private 1557 356,571
Public 461 188,090
Panel C: Distribution of Green Bonds by Industry
N Total (MM)
Bank 866 240,350
Financial 162 62,091
Industrial 933 227,183
Insurance 7 2,563
Municipal 2 945
Real Estate 195 20,101
Utility 30 15,495
Panel D: Distribution of Green Bonds by Country
N Total (MM)
China 275 109,085
France 193 39,585
Italy 21 10,267
Japan 69 10,762
Mexico 9 12,186
Netherlands 81 53,496
Norway 34 8,188
Others 578 159,694
SNAT 445 85,766
Sweden 220 18,548
UK 22 8,005
US 248 53,147
Panel E: Summary statistics of key variables
mean sd min p25 p50 p75 max
Green bonds (%) 17.009 24.179 0.070 0.742 5.043 21.332 92.589
Carbon price ($) 32.480 37.202 0.882 8.042 11.364 55.519 163.147
Environmental score 48.998 23.079 0.000 29.480 47.340 67.560 99.310
Firm Size ($B) 239.843 5,422.056 0.000 0.102 0.641 3.981 4.48e+05
Scaled WPS 529.688 15771.873 0.992 7.096 13.615 40.943 8.69e+05
Share turnover 142.685 407.876 0.000 71.814 97.905 135.681 7,880.690
Firm CO2 emissions (Mt) 4.00 10.8 0.000 0.082 0.359 2.14 99
Moreover, our proxies WPS and Turnover both vary significantly across industries.
The ANOVA test rejects the hypothesis that industries’ mean WPS and Turnover are
equal at the 1% level.
Replication and additional robustness exercises
The following estimation results complete those presented in Table I.
Table V shows how the results of Section IV are modified when standard errors are
clustered at the country-industry level. When Incentives=WPS, the coefficient of the
interaction component of (E.1) becomes significantly different from zero at the 11% level.
The difference with results presented in Table I raises the possibility that an unobserved
variable affect green bond issuance at the industry level.
Moreover, Table VI shows that our results do not change qualitatively when the
managerial stock price sensitivity parameter of our model is not redefined by ˜
α/(2 α) as per Corollary 1.
We now examine the total effect of managerial incentives according to the regression
results presented in Section IV, Table I. According to these results, the total effect is
Table V: Green bond issuance, managerial incentives, and carbon price:
alternative clustered standard errors
This table presents estimates from panel regressions of the proportion of green bonds issued by firms on
their countries’ effective carbon prices and proxies of managers’ stock-price sensitivity in their industries.
We control for firm size, book-to-market ratio, and environmental score. Standard errors are clustered
at the country-industry level. *, **, and *** denote significance at the 10%, 5%, and 1%, respectively.
Green bonds
(1) (2) (3) (4)
Incentives (WPS) -0.505**
Carbon price ×Incentives (WPS) 0.017 0.024
(0.010) (0.021)
Incentives (Turnover) -0.158***
Carbon price ×Incentives (Turnover) 0.006*** 0.005**
(0.002) (0.002)
Controls Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes
Industry FE Yes No Yes No
Industry-year FE No Yes No Yes
Country-year FE Yes Yes Yes Yes
Observations 15011 15008 15148 15145
R20.335 0.358 0.335 0.358
Table VI: Green bond issuance, managerial invcentives, and carbon price:
simplifying Incentives variable redefinition
This table presents estimates from panel regressions of the proportion of green bonds issued by firms on
their countries’ effective carbon prices and proxies of managers’ stock-price sensitivity in their industries,
where the latter are not redefined as per Corollary 1. We control for firm size, book-to-market ratio,
and environmental score. Standard errors are clustered at the country level. *, **, and *** denote
significance at the 10%, 5%, and 1%, respectively.
Green bonds
(1) (2) (3) (4)
Incentives (WPS) -0.953***
Carbon price ×Incentives (WPS) 0.031*** 0.045*
(0.010) (0.024)
Incentives (Turnover) -0.147***
Carbon price ×Incentives (Turnover) 0.006*** 0.005***
(0.002) (0.002)
Controls Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes
Industry FE Yes No Yes No
Industry-year FE No Yes No Yes
Country-year FE Yes Yes Yes Yes
Observations 15011 15008 15148 15145
R20.335 0.358 0.335 0.358
Capital GoodsCapital GoodsCapital Goods
0 .005 .01
Green bonds
.06 .08 .1 .12 .14 .16
Incentives (Turnover)
Figure VI: Certified green bond issuance and stock share turnover (2007-2019)
positive on average, i.e., at least for any effective carbon price greater than the OECD
average of 32. Yet, this total effect is not statistically different from zero at the 32
In order to test whether the total effect of managerial incentives is statistically different
from zero at a given threshold level Carbon price, we rewrite the empirical model (E.1)
by reducing countries’ effective carbon prices by this threshold level, which yields:
Green bondsi,t =β0+η1Incentivesj(i),t1
+η2Carbon pricec(i),t1Carbon price×Incentivesj(i),t1
+β3Controlsi,t1+Fixed effects +i,t,(E.1)
where η1β1×Carbon price +β2becomes the coefficient of the total contribution of
managerial incentives.
Consider, for example, the average effective carbon price in the EU, accounting for
about 50% of the global volume of green bonds: Carbon price = $81.75. Table VII shows
our regression results with this threshold. The coefficient of the total contribution of
managerial incentives is statistically different from zero at the 10% level.
Table VIII presents various regressions that extend those of Section IV. It examines the
firm-level variation in stock share turnover by using Turnoveri,t rather than the industry-
level aggregate Turnoverj(i),t. Moreover, it includes firms’ Foreign sales,Foreign assets,
and Foreign income, on top of firm fixed effects, to deal with multinationals.
Table VII: Green bond issuance, managerial incentives, and carbon price:
total contribution at the EU average
This table presents estimates from panel regressions of the proportion of green bonds issued by firms on
their countries’ effective carbon prices reduced by the average EU carbon price and proxies of managers’
stock-price sensitivity in their industries. We control for firm size, book-to-market ratio, and environ-
mental score. Standard errors are clustered at the country level. *, **, and *** denote significance at
the 10%, 5%, and 1%, respectively.
Green bonds
(1) (2) (3) (4)
Incentives (WPS) 0.908*
Carbon price ×Incentives (WPS) 0.017*** 0.024*
(0.006) (0.014)
Incentives (Turnover) 0.340**
Carbon price ×Incentives (Turnover) 0.006** 0.005***
(0.002) (0.002)
Controls Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes
Industry FE Yes No Yes No
Industry-year FE No Yes No Yes
Country-year FE Yes Yes Yes Yes
Observations 15011 15008 15148 15145
R20.335 0.358 0.335 0.358
Table VIII: Green bonds issuance, carbon tax, and short-termism: robustness
This table presents estimates from panel regressions of the proportion of green bonds issued by firms on
their countries’ effective carbon prices and proxies of managers’ stock-price sensitivity. We control for
firm size, book-to-market ratio, environmental score, and measures of firms’ foreign activities. Standard
errors are clustered at the country level. *, **, and *** denote significance at the 10%, 5%, and 1%,
Green bonds
(1) (2) (3) (4)
Carbon price ×Incentives (Firm-level Turnover) 1.463** 1.588** 0.403* 0.441*
(0.637) (0.682) (0.235) (0.240)
Foreign sales 0.250
Foreign assets 229.362*
Foreign income 2.191
Controls Yes Yes Yes Yes
Firm FE Yes Yes Yes Yes
Industry-year FE Yes Yes Yes Yes
Country-year FE Yes Yes Yes Yes
Observations 15145 14008 11912 11260
R20.358 0.359 0.428 0.429
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
We show that executives cut investment when their incentives become more short term. We examine a unique event in which hundreds of firms eliminated option vesting periods to avoid a drop in income under accounting rule FAS 123-R. This event allowed executives to exercise options earlier and thus profit from boosting short-term performance. Our identification exploits that FAS 123-R’s adoption was staggered almost randomly by firms’ fiscal year-ends. CEOs cut investment and reported higher short-term earnings after option acceleration, and they subsequently increased equity sales.
Macro-finance analyses commonly link firms’ borrowing constraints to the liquidation value of physical assets. For U.S. nonfinancial firms, we show that 20% of debt by value is based on such assets (asset-based lending in creditor parlance), whereas 80% is based predominantly on cash flows from firms’ operations (cash flow–based lending). A standard borrowing constraint restricts total debt as a function of cash flows measured using operating earnings (earnings-based borrowing constraints). These features shape firm outcomes on the margin: first, cash flows in the form of operating earnings can directly relax borrowing constraints; second, firms are less vulnerable to collateral damage from asset price declines, and fire sale amplification may be mitigated. Taken together, our findings point to new venues for modeling firms’ borrowing constraints in macro-finance studies.
Assets in the fossil fuel industries are at risk of losing market value due to unanticipated breakthroughs in renewable technology and governments stepping up climate policies in light of the Paris commitments to limit global warming to 1.5 or 2°C. Stranded assets arise due to uncertainty about the future timing of these two types of events and substantial intertemporal and intersectoral investment adjustment costs. Stranding of assets mostly affects the 20 biggest oil, gas, and coal companies who have been responsible for at least one-third of global warming since 1965, but it also affects carbon-intensive industries such as steel, aluminum, cement, plastics, and greenhouse horticulture. A disorderly transition to the carbon-free economy will lead to stranded assets and legal claims. Institutional investors should be aware of these financial risks. A broader definition of stranded assets also includes countries reliant on fossil fuel exports and workers with technology-specific skills. Expected final online publication date for the Annual Review of Resource Economics, Volume 12 is October 5, 2020. Please see for revised estimates.
Cambridge Core - Finance and Accountancy - Grow the Pie - by Alex Edmans
We document that an increase in short-horizon investors is associated with cuts to long-term investment and increased short-term earnings. This leads to temporary boosts in equity valuations that reverse over time. To estimate these effects, we use difference-in-differences regressions around firms’ additions to the Russell 2000, comparing firms with large and small increases in short-term ownership. We proxy for the presence of short-term investors using ownership by transient institutions. Our results suggest that short-term pressures by investors can lead to myopic firm behavior. This paper was accepted by Shiva Rajgopal, accounting.
A structural model for green bonds is developed to explain the formation and dynamics of green bond prices and to address the issue of the so-called ‘greenium’, that is, the difference between the yields on a conventional bond and a green bond with the same characteristics. We provide answers to the following questions: What are the determinants of the green bond value? Do green bonds enhance the credit quality of the issuer? Are green bonds a relatively cheap tool to fund sustainable investments? We also study the effect of investors' environmental concern on portfolio allocation. Our results have direct policy implications and suggest that an improvement in credit quality could ultimately lead to a lower cost of capital for green bond issuers and that governmental tax-based incentives and an increase in investors' green awareness play a significant role in scaling up the green bonds market.
We investigate the association of new housing construction with sea level rise exposure and climate change beliefs. We combine U.S. Census construction permits data, sea level rise projections, county-level climate change belief estimates (Howe et al., 2015), and standard housing controls. The results suggest that (i) sea level rise vulnerability is associated with significantly reduced construction in areas with high climate change beliefs, but that (ii) this relationship is significantly attenuated in more skeptical areas, suggesting that climate skepticism may be delaying adaptation.