2. The model

Let us consider a firm with assets-in-place that generate uncertain earnings before interest and taxes (EBIT) described by a stochastic process of the form:

(1)$$dY_t =\mu Y_t dt+\sigma Y_t dW_t\comma \; $$

where W _t is a standard Wiener process with respect to an assigned filtration {I _t}_t≥0. The size of Y _t is determined by the amount of the firm's output and this random variable is affected by the demand shock for the firm's product as well.

Let r be the market risk-free interest rate and let the growth rate be such that μ < r, as is usual in the financial literature on corporate defaultable debt, where this assumption is adopted for convergence reasons. Let τ denote the corporate tax rate that the firm faces on its income after servicing possible interest payments on its debt.

The production process generates environmental damage, because the employed technology is supposed to be ‘brown’. The environmental damage caused will eventually result in a proportional reduction in the firm earnings.Footnote ⁴ This can take the form of a fee that is paid by the firm in relation to its carbon emissions or any other form of decrease (e.g., due to reputational issues, or allegations concerning negative impact on the ecosystem) in the firm earnings. Let D denote the damage per unit of production and let p denote the proportional reduction of the firm's income. We can set p ∈  [0, 1], where p < 1 means that not all damage is internalized by the firm, either because it cannot be wholly intercepted or because the social impact is underweighted.

The firm may decide to go ‘green’ and exercise an investment option to develop a cleaner technology. This mitigation policy will result in a smaller environmental damage and thus in a substantial reduction in the penalty on earnings. Let g > 0 denote the intensity of the investment expenditure in the greener technology and let δ be a parameter representing the effectiveness of the technology. We suppose that the green investment allows the initial damage, D ₀, to be reduced by a factor e ^{−δ g}. Denote D _g = e ^−δgD ₀. Before investing in the new technology, EBIT are penalised by an amount pY _t D ₀, while upon exercising the green investment option the reduction is only pY _t D _g. A discussion on how the analysis would be affected through non-multiplicative damages is offered in remark 2, where the case of a damage that is proportional to a power of Y (Gollier, Reference Gollier2018) is addressed. Finally, one could incorporate possible random fluctuations of the variable D and specify the stochastic process, for example, as in van den Bremer and van der Ploeg (Reference van den Bremer and van der Ploeg2018). Although we assume that the exercise of the green option is irreversible, the firm has the flexibility to exercise the option at any time.

The main assumption in this model is that the firm will finance its green policy by issuing a perpetual green bond with a continuous coupon. The management chooses to exercise the green option in order to maximize the market value of equity. We do not model conflicts of interest between management and equity holders by assuming that the firm is managed by equity holders and that the equity holders also have the option to stop operations and trigger default in an optimal way. Let Y _I denote the level of the fundamental variable Y at which it is optimal to issue a green bond, and let Y* denote the default threshold, i.e., default is triggered whenever Y falls below Y*. In case Y falls below Y*, then the firm goes bankrupt and the original debt holders take over and obtain the firm's unlevered assets net of proportional bankruptcy costs, α(0 ≤ α ≤ 1).The following requirements should be satisfied to have a green bond:

1. (i) all the proceeds from the bond issuance have to be used in the green project;

2. (ii) special reporting, monitoring and accounting are needed to guarantee accomplishment of a Green Standard.

We model (i) by assuming that the investment amount equals the bond value at the investment threshold, and we add a fixed cost, K, representing the extra costs related to the requirements in (ii), including the cost of setting up specific teams fully dedicated to the engagement with sustainable goals or the expenditures for external assessment of the sustainability credentials and management of the green bond.

3. Green bond pricing

Let G denote the current value of the green bond. The bond is issued at the investment time and the contractual continuous coupon, g, is tax deductible. If the value of the fundamental variable Y is above the default threshold Y*, then G(Y) satisfies the following equation:

(2)$$\displaystyle{{\sigma ^2}\over{2}}Y^2\partial _Y^2 G+\mu Y\partial _Y G-rG+g=0\comma \; $$

where r is the risk-free interest rate. A trivial solution of (2) is G(Y) = g/r, which represents the value of a default-free bond. If we look for a general solution of the homogeneous equation related to (2) by guessing a form Y ^β, then β needs to solve the following equation:

$$\displaystyle{{\sigma ^2}\over{2}}\beta \lpar \beta -1\rpar +\mu \beta -r=0\comma \;$$

that is,

$$\beta _\pm =\displaystyle{{1}\over{2}}-\displaystyle{{\mu }\over{\sigma ^2}}\pm \sqrt {\left({\displaystyle{{1}\over{2}}-\displaystyle{{\mu }\over{\sigma^2}}} \right)^2+\displaystyle{{2r}\over{\sigma ^2}}}.$$

Upon default, bond holders receive the value of the firm (evaluated at Y*) minus bankruptcy costs amounting to the fraction $\alpha \lpar 0 \leq\alpha \leq1\rpar $ of firm value. Combining the default value of G with the limit (or no-bubble) condition G(∞) = g/r, which allows us to get the riskless bond value g/r for large values of Y, we get the following expression for the bond value:

(3)$$G\lpar Y\rpar =\displaystyle{{g}\over{r}}+\left[{\lpar 1-\alpha\rpar P_g Y^{\ast}-\displaystyle{{g}\over{r}}}\right]\left({\displaystyle{{Y}\over{Y^{\ast}}}}\right)^{\beta_{-}},$$

where P _g is a short notation for 1 − τ − pD _g/r − μ. Note that β ₋ < 0, while β ₊ > 1, and the shape of (3) is obtained by adopting the no-bubble condition above for large values of Y, as is usual in the literature on corporate bonds.

Default is triggered by equity holders; we therefore compute the default threshold as the level of Y maximizing the post-investment equity value, E(Y). The post-investment equity value solves the following equation:

(4)$$\displaystyle{{\sigma ^2}\over{2}}Y^2\partial _Y^2 E+\mu Y\partial _Y E-rE+\lsqb Y-g\rsqb \lpar 1-\tau\rpar -pD_g Y=0,$$

and the following boundary conditions:

$$E\lpar {Y^\ast } \rpar =0\comma \; E'\lpar {Y^\ast } \rpar =0\comma \;$$

ensuring that the equity value equals zero at bankruptcy, due to limited liability, and that equity holders choose the bankruptcy threshold optimally. If the no-bubble condition lim _Y→∞ E(Y)/Y < ∞ is adopted, then E(Y) has the following expression:

(5)$$E\lpar Y\rpar =P_g Y-\displaystyle{{g\lpar 1-\tau \rpar }\over{r}}+\left[{\displaystyle{{g\lpar 1-\tau \rpar }\over{r}}-P_g Y^\ast } \right]\left({\displaystyle{{Y}\over{Y^\ast}}} \right)^{\beta_{-}}\comma \; $$

where

(6)$$Y^\ast =\displaystyle{{\beta_{-}}\over{\beta_{-}-1}}\displaystyle{{g\lpar 1-\tau \rpar }\over{rP_g}}. $$

Remark 1: As D _g < D ₀, expression (6) predicts a lower default threshold for a green bond than for a conventional bond with otherwise equal indentures, that is, default is delayed with green bonds. Combining (3) and (6), one gets a higher bond value for the green issuance for fixed parameter values. This effect is confirmed through the numerical computation of section 4 where, additionally, the issue of the cost of the green labelling is addressed as well. The total firm value after investment is the sum of equity and debt values:

(7)$$V\lpar Y\rpar =P_g Y+\displaystyle{{g\tau}\over{r}}-\left[{\displaystyle{{g\tau}\over{r}}+\alpha P_g Y^\ast}\right]\left({\displaystyle{{Y}\over{Y^\ast}}} \right)^{\beta_{-}}. $$

Equity holders choose the investment threshold Y _I to maximize the ex ante equity value, V ₀ (Y). This is of the form $P_{0} Y+\hat{A}Y^{\beta _+}$, where the latter represents the investment option.Footnote ⁵ By matching the ex ante and ex post values, net of the cost K, and their derivatives at the point Y _I, the investment threshold and the arbitrary constant  can be determined. The extra cost K embodies the additional expenditures related to a green bond issuance, including separate accounting, additional monitoring and reporting, as requested by the GBPs. The two matching equations are solved numerically for  and Y _I.

Remark 2: In Gollier (Reference Gollier2018) the damage function is assumed to be proportional to Y ^ξ where Y is the per-damage output. So one may wonder how our model is changed if a multiplicative damage function (ξ = 1) is replaced with a power function.Footnote ⁶ Extending our arguments in this new framework, we find that the equation (6) for the default threshold should be replaced with the following equation:

$$\displaystyle{{1-\tau}\over{r-\mu}}Y^{\ast}+\displaystyle{{pD_g}\over{0.5\xi \lpar \xi -1\rpar +\mu \xi -r}}\lpar Y^{\ast}\rpar ^\xi +\displaystyle{{\beta_{-}}\over{1-\beta_{-}}}\displaystyle{{g\lpar 1-\tau\rpar }\over{r}}=0.$$

As a general result is hardly obtained, let us focus on the case ξ ≈ 1 and μ < σ ²/2. In this case, it can be shown that when $\xi >1 \ \lpar \xi <1\rpar $ the default threshold is decreased (increased) with respect to the linear case, and the greenium is increased (decreased) relative to the linear case. Thus, our results are amplified in the case of damages proportional to Y ^ξ with ξ > 1.

Finally, we compute the optimal level of the coupon, g*(Y), under a ‘first-best’ scenario, that is, when a firm-value-maximizing investment policy is pursued. Specifically, one gets the optimal coupon level by solving the equation ∂_g V(Y) = 0 for g, where the firm value is given by (7). The equation can be written as follows in a more explicit form:

(8)$$\displaystyle{{p\delta gD_0}\over{\lpar 1-\tau \rpar e^{\delta g}-pD_0}}\lsqb \lpar \beta^{-1}\rpar \tau +\alpha \lpar 1-\tau \rpar \rsqb +\lpar 1-\beta^{-}\rpar \tau -\left({\displaystyle{{Y^{\ast}}\over{Y}}}\right)^{\beta^{-}}=0. $$

Of course, the optimal coupon level can be obtained from equation (8) through numerical computation only.

4. Numerical computation

In what follows we perform numerical computation to compare the green bond value and yield at issuance to an otherwise comparable non-green bond. As an illustration, we consider a green bond (ISIN Code: XS1550149204) issued by Enel, the main Italian utility producing and distributing electricity. Enel committed to decarbonize its energy mix and to contribute to achieving four of the seventeen UN Sustainable Development Goals, in particular SDG 13. The issuance of green bonds is seen as an ideal tool to finance the transition to a low carbon economy. The bond was issued on 16 January 2017 with a coupon rate of 1 per cent and will expire on 16 September 2024. Our sample size consists of 338 observations and the parametric model by Nelson and Siegel has been used to build the yield curve for the issuer. The analysis confirms a greenium with median 6.89 basis points (bps) and a range between 0.43 and 17.96 bps.

Our theoretical model is implemented by using the following parameter values: r = 0.385 per cent is extracted from the EU yield curve at the time of our observations by interpolation, μ = −0.286 per cent and σ = 28.86 per cent are estimated from the Enel stock prices in the year before the bond issuance. The Italian corporate tax rate is 27.9 per cent including IRES (Imposta sul REddito delle Società – corporate income tax) and IRAP (Imposta Regionale sulle Attività Produttive – regional tax on production). Bankruptcy costs are assumed at the 15 per cent level, in keeping with empirical literature in corporate finance (e.g., Altman and Hotchkiss, Reference Altman and Hotchkiss2006). The parameters relating to the green projects are calibrated in order to achieve an advantage of 20 per cent to sustainable returns, as empirical estimates report that, on average, the market ascribes a valuation premium to sustainable competitive advantage that tends to be capped out at this level (Ling et al., Reference Ling, Forrest, Fox and Feilhance2007). We compute the greenium, where the greenium is defined as the difference between the yields on a conventional bond and a green bond with the same characteristics. If the issuance cost for the green bond, K, is fixed at the 2 per cent level of the earnings, we get a greenium of 4.17 bps; if the cost is larger, the green bond issuance is delayed and the greenium at issuance is slightly larger. On the other hand, if a periodic cost is subtracted from earnings, the greenium is reduced (for example, to 3.82 bps with a 1 per cent annual cost) as the negative effect on earnings materializes through a premium required by the market; in any case, whenever the costs are not too high, a significant positive premium to the issuer can be detected. Note that the greenium we get in the base case is included in the range of the greenium we found in the empirical analysis, but it is slightly smaller than the estimated mean and median values. An improvement might come from a more precise estimation of the costs and other considerations that are not modelled in this framework, for example, possible special investment allowances or guarantees to bond holders.Footnote ⁷ On the other hand, it is known that structural models usually generate lower credit spreads than those estimated from empirical data, which comes from the implicit assumption of complete information available to market players. A better fit of market data could be obtained by adopting an incomplete knowledge of the firm's fundamentals, as suggested in the financial literature on incomplete information.

Incidentally, we can compute that with these parameter values the default threshold for the green bond is decreased relative to the default threshold of a comparable conventional bond. Figure 1 depicts the default trigger for a green and a conventional bond as functions of the coupon level. As expected, default occurs earlier for higher values of the interest payment, but the default threshold is lower if the bond proceeds contribute to better sustainable returns for the firm, as is the case with green investments. An improvement in credit quality could ultimately lead to a lower cost of capital for green bond issuers.

Figure 1.

Default thresholds as a function of the coupon level

Let us perform sensitivity analysis with respect to the relevant parameters, while keeping the other parameters the same as in the base case. As table 1a shows, a higher asset volatility increases the greenium as the default threshold is lowered. The effect of asset volatility on the default trigger is a common feature in many structural models in corporate finance and, as a result of the shift in the default threshold, a higher volatility decreases the risk premium on a bond yield. However, the impact on the greenium is not straightforward. As for the parameters governing the green technology, an improvement related to a more effective technology widens the greenium, as expected. Similarly, the higher the penalty is for not investing in a green project, and therefore the larger the sustainability advantage, the wider the greenium becomes. On the other hand, a decrease in the corporate tax rate increases earnings, but reduces the tax benefit of the debt service, and the net effect shrinks the green premium to the issuer.

Table 1a.

Sensitivity of the greenium to asset volatility (σ)

Table 1b.

Sensitivity of the greenium to the effectiveness of the green technology (δ)

Table 1c.

Sensitivity of the greenium to the sustainability advantage/disadvantage (p)

Table 1d.

Sensitivity of the greenium to the corporate tax rate (τ)

5. Why do investors choose green bonds?

In this section we model the demand side of a green bond by solving the problem of a prototypical (representative) consumer-investor who can invest both in a green bond and in a conventional bond issued by the same company. Following a stream of economic literature on environmental risk, we assume that damages affect future consumption multiplicatively (see Lemoine, Reference Lemoine2017); that is, the instantaneous utility, U, is a function of c/F(D), where c is the consumption rate and F is a function of the damage. In the models of climate change, F(D) is eventually expressed as a function of global warming. Here we adopt a simple expression to keep the setting very generic and assume that F(D) = 1 + εD, where ε…0 is a measure of the consumer's concern about environmental risk, for example, the case ε = 0 represents an environmentally neutral individual.

Let N _g denote the number of green bonds and N _b the number of conventional bonds in the investor's portfolio. Let R(t) denote the total wealth at time t. Then the consumer-investor's budget equation is:

(9)$$dR=N_g \lpar t\rpar dG\lpar t\rpar +N_b \lpar t\rpar dB\lpar t\rpar -c\lpar t\rpar dt\comma \; $$

where G (respectively, B) is the green (respectively, conventional) bond value that has been obtained in section 3 in explicit form. Let x = N _g G/R denote the share of wealth in the green bond and thus 1 − x = N _b B/R is the share of wealth invested in the conventional bond. In view of expression (3) for the green bond value G and a similar expression for B, the budget equation can be rewritten as:

(10)$$dR=\displaystyle{{\beta^{-}R}\over{r}}\left[{r-\displaystyle{{b}\over{B}}+x\left({\displaystyle{{b}\over{B}}-\displaystyle{{g}\over{G}}}\right)}\right]\displaystyle{{dY}\over{Y}}-cdt. $$

For notational convenience let us denote b/B by $\hat{b}$ and g/G by ĝ, that is, $\hat{b}$ and ĝ are the yields on the two bonds. In what follows, the yields will be evaluated at the issuance, although a more general case can be handled within this framework. Finally, as dY _t = μY _t dt + σY _t dW _t, we get the following budget equation:

(11)$$dR=\left[{\displaystyle{{\beta^{-}\mu R}\over{r}}\lpar r-x{\hat g}+\lpar x-1\rpar {\hat b}\rpar -c} \right]dt+\left[{\displaystyle{{\beta^{-}\sigma R}\over{r}}\lpar r-x{\hat g}+\lpar x-1\rpar {\hat b}\rpar } \right]dW_t. $$

The consumer-investor's problem is:

(12)$$\hbox{max}_{x\comma c} \; E=\left[{\vint_0^\infty {e^{-rt}U\left({\displaystyle{{c_t}\over{1+\varepsilon D}}} \right)dt}} \right]\comma \; $$

such that the state variable R follows (11) and Ro is given. Denote the current value of the value function by J(R). Then the Bellman equation is:

$$\eqalign{&\hbox{max}_{{\rm x\comma c}} \left[U\left({\displaystyle{{c}\over{1+\varepsilon D}}} \right)-rJ\lpar R\rpar +\left[{\displaystyle{{\beta^{-}\mu R}\over{r}}\lpar r-x{\hat g}+\lpar x-1\rpar {\hat b}\rpar -c} \right]J^{\prime}\lpar R\rpar \right. \cr & \quad +\left. \displaystyle{{\sigma^2}\over{2}}\left[{\displaystyle{{\beta^{-}R}\over{r}}\lpar r-x{\hat g}+\lpar x-1\rpar {\hat b}\rpar }\right]^2J{\prime\prime}\lpar R\rpar \right]=0.}$$

The first-order conditions yield:

(13)$$U{\prime} \left({\displaystyle{{c}\over{1+\varepsilon D}}}\right)=\lpar 1+\varepsilon D\rpar J{\prime} \lpar R\rpar \comma \; $$

(14)$$x=\displaystyle{{r-{\hat b}}\over{{\hat g}-{\hat b}}}+\displaystyle{{\mu r}\over{o^{2}\beta ^{-}\lpar {\hat g}-{\hat b}\rpar }}\displaystyle{{J{\prime}\lpar R\rpar }\over{RJ^{\prime\prime}\lpar R\rpar }}. $$

Note that the values of $\beta^{-}\comma \; {\hat b}\comma \; {\hat g}$ depend on the parameters r, μ, σ, τ, α and therefore the behaviour of x with respect to the relevant parameter is hard to guess. In order to gain insight, we explicitly solve the problem in the case of an exponential utility function, U(C) = −exp( − γC)/γ, where γ embodies risk aversion. Note that this utility function does not separate risk aversion and intertemporal substitution, as, for example, Epstein-Zin preferences.

To obtain an explicit expression for the value function, we look for a solution such that $J' \lpar R\rpar =A\cdot \exp \lpar a\sqrt R\rpar $, with A and a to be determined. Solving (13) for c and plugging the expression obtained for c and (14) into the Bellman equation above, we get $a=-\sqrt {4\gamma \varphi /\lpar 1+\varepsilon D\rpar } $ and A = 1/1 + εDexp(1 − r/2φ) with φ = r/2 + (μ/2σ) ². Thus, the optimal share of green bonds in the portfolio, when the wealth R is normalized to 1, becomes:

(15)$$x=\displaystyle{{{\hat b}-r}\over{{\hat b}-{\hat g}}}+\displaystyle{{\mu r}\over{o^2\beta ^-\lpar {\hat b}-{\hat g}\rpar }}\displaystyle{{\sqrt {1+\varepsilon D} }\over{\sqrt {\gamma \varphi}}}. $$

Expression (15) specifies the optimal allocation between green bonds (x) and conventional bonds (1 − x). Note that the share of green bonds is at least as large as the share of conventional bonds whenever

(16)$${\hat b}-{\hat g} \lt {\hat g}-r+\displaystyle{{\mu r}\over{o^{2}\beta^{-}}}\displaystyle{{\sqrt {1+\varepsilon D} }\over{\sqrt {\gamma \varphi } }}\comma \; $$

that is, the ‘greenium’ $\lpar {\hat b}-{\hat g}\rpar $ should not exceed the credit spread on the green bond $\lpar {\hat g}-r\rpar $ augmented by a premium that increases with ε and decreases with the risk aversion parameter γ.

As an exemplification, in figure 2 we show how the portfolio allocation can be affected by the degree of a bond holder's concern about environmental damage (embodied by ε). We use annualized returns on the S&P Dow Jones Green Bond Index (DJ-GB) and Aggregate Bond Index as of 31 October 2018. Green bond indices are a good indicator to analyze the secondary market performance of green bonds from an investor's perspective, as suggested by Ehlers and Packer (Reference Ehlers and Packer2017). In fact, they contain a diversified broader portfolio of bonds and therefore provide a good means of comparison with the performance of other bond indices for a wide range of investors. Also Reboredo (Reference Reboredo2018) uses the S&P Dow Jones Green Bond Index to track the financial performance of green bonds. Figure 2 shows the sensitivity of the two shares x and (1 − x) to the parameter ε. We set both portfolio weights at 50 per cent for ε = 0 by selecting the risk-aversion parameter γ, that is, we start with environmentally ‘neutral’ investors who allocate their wealth between the two bonds to diversify financial risk. ε is then increased to represent a higher degree of green commitment. For very large values of ε, the bond holder overinvests in the green bond, even by taking a short position in the other financial security.

Figure 2.

Asset allocation as a function of environmental preferences

As a consequence, a widespread environmental awareness will lead to an increased demand for green bonds and to issuing oversubscriptions, which would provide a cheaper way to fund environmentally- sustainable projects.

This analysis complements the stream of research on the effect of green investing on the corporate behavior of firms (see, e.g., Heinkel et al., Reference Heinkel, Kraus and Zechner2001), where the cost of capital is obtained from the relative proportion of green versus non-green investors. However, in our paper the focus is on bond investors and the dynamics refers to the firm policy instead of market equilibria.

The value consumers-investors attach to environmentally-sustainable products, and thus a measure of their green mandate, could be extracted by some empirical analyses, or field experiments. A few recent papers (Hainmuller et al., Reference Hainmuller, Hiscox and Sequiera2015; Becchetti et al., Reference Becchetti, Salustri and Scaramozzino2018) provide evidence about the values consumers attach to ethical sourcing. They applied various treatments to a wide range of green products and tested the impact of a simple form of advertising on environmentally responsible products with and without increase in prices, to show that public information on corporate environmental goals generates significant changes in demand for these products.

In order to provide evidence of a link between asset allocation and an investors' green mandate, one needs data on the percentage of the total green debt that is sold to green investors. Some 2017 joint reporting by the Climate Bonds Initiative and the International Finance Corporation provides evidence of a significant portion allocated to those labeling themselves as socially responsible, or green investors, with an average allocation of 62 per cent in the developed markets and of 25 per cent for emerging markets. Overall, the percentage of green bonds allocated to buyers who identify themselves as green investors is increasing and is expected to increase for emerging markets as well.

Finally, we stress that the disclosure of investors' preferences toward environmental sustainability has been recently integrated into the suitability obligations arising from EU Directive 2014/65 (MiFID II) and EU Directive 2016/97 (IDD). The proposal, released on 24 May 2018, recommends the introduction of questions that would help investment firms and insurance intermediaries identify the clients' investment objectives, including environmental preferences. Therefore, each client's individual environmental, social and governance (ESG) concern will be collected and considered in the design of the financial products offered and aggregate information regarding the investors' preferences for environmentally responsible investing might soon be disclosed.

6. Tax advantage

The promotion of green lending sometimes takes the form of tax-based incentives. Central and local governments can take measures to attract capital from investors and expand the green bond market. These forms of supporting policies can be helpful especially in the earliest stage, when investors need to gain confidence in the new financial tools and the issuers are developing a credit history in the green sector. Tax incentives for bonds financing renewable energy and green buildings are present in the US and have been proposed for labeled green bonds in various countries. In India, some concessions in the form of tax exemptions were introduced to stimulate the domestic demand for green bonds. The tax-free bond issued by the Indian Renewable Energy Development Agency Limited (IREDA) in 2016 was oversubscribed by more than five times.

In this section we study the impact of a favorable tax rate on the proceeds from a green bond in comparison to a conventional bond. We fix the tax rate for conventional bonds at 25 per cent, which is a common percentage across Europe, while varying the tax rate that bondholders have to pay on the coupons of a green bond. We assume that the bond holder equally splits his/her portfolio between green and non-green bonds when the tax rate is the same (25 per cent) for both bonds.

As figure 3 shows, the tax rate on the coupon has a significant impact on the investors' decision. Thus governmental tax-based incentives play a significant role in scaling up green bond markets.

Figure 3.

Bond portfolio vs tax rate on green bonds