Economic Modelling xxx (xxxx) xxx

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.journals.elsevier.com/economic-modelling

Price connectedness between green bond and financial markets

Juan C. Reboredo a, *, Andrea Ugolini b
a
Department of Economics, Universidade de Santiago de Compostela, Spain
b
Department of Quantitative Analysis, Universidade do Estado do Rio de Janeiro, Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: We study price connectedness between the green bond and financial markets using a structural vector autore-
Green bonds gressive (VAR) model that captures direct and indirect transmission of financial shocks across markets. Using
Financial markets heteroskedasticity to identify the structural VAR model parameters, our empirical findings reveal that the green
Price spillovers
bond market is closely linked to the fixed-income and currency markets, receiving sizeable price spillovers from
Structural VAR
those markets and transmitting negligible reverse effects. We also show that, in contrast, the green bond market is
weakly tied to the stock, energy and high-yield corporate bond markets. These findings have implications in terms
of portfolio and risk management decisions for environmentally aware investors holding positions in green bonds.

1. Introduction Initiative, 2018). Even so, the green bond market still accounts for only a
relatively small fraction (around 17%) of unlabelled climate-aligned
Scaling up the transition to a climate-resilient economy requires bonds and an even smaller share (less than 1%) of the bonds market.
mobilizing financial resources for green projects and divestment in fossil Green bonds are mostly issued in USD and EUR (around 80% of issu-
energies (OECD, 2016), which would open up new green finance op- ances) by private companies, the public sector (national government,
portunities for individual and institutional investors. Because they local government or state entities) and supranational entities (e.g., the
channel financial resources to environment-friendly projects, green World Bank and the European Investment Bank) and have an average
bonds are becoming increasingly popular and established financial in- tenor of 5–10 years. The proceeds of green bonds are used to fund
struments within the sustainability-oriented financial community. renewable energy projects, energy efficiency projects, low-carbon
Green bonds have features similar to conventional fixed-income transport, sustainable water projects and waste and pollution projects,
corporate bonds, except that their proceeds are earmarked for environ- accounting in 2015 for 45.8%, 19.6%, 13.4%, 9.3% and 5.6%, respec-
mentally friendly projects. The integrity and transparency of the green tively, of issuances (European Commission, 2016).
bond market has been boosted by publication, in January 2014, of the In the light of investors’ interest in environmental risk and the 2015
Green Bond Principles (GBP) by the International Capital Markets As- Paris Agreement — in which a wide set of countries pledged transition to
sociation (ICMA). The GBP set global standardized rules for the labelling a climate-resilient economy — the green bond market is expected to
of a bond as green, bearing in mind the environmental impact of its thrive. In fact, green bonds are recognized as an appropriate instrument
proceeds. Furthermore, the opening of specific green bond segments in to finance the transition to a low-carbon economy (OECD, 2017; Mon-
stock exchanges — such as those in Italy, Oslo, London, Mexico, asterolo and Raberto, 2018) and have the capacity to redistribute the cost
Luxembourg, Shanghai and Shenzhen — has contributed to both of mitigating climate change across generations (Flaherty et al., 2016).
improving the liquidity and reputation of green bonds as a distinctive However, the rate at which the green bond market develops closely de-
financial instrument and enhancing their attractiveness for both issuers pends on the risk-return profile of green bonds and also on how their
and individual and institutional investors. prices may be impacted by price oscillations occurring in other financial
Since the first green bond issue by the European Investment Bank in markets. Understanding price transmission between green bond and
2007, issuances in the green bond market have grown from 11bn USD in other financial markets is crucial to determining the performance of
2013 to 37bn USD in 2014, 43bn USD in 2015, 81bn USD in 2016, green bonds and their usefulness for hedging and managing portfolio
155.5bn USD in 2017 and 167.6bn USD in 2018 (Climate Bonds risks (Reboredo, 2018), as these ultimately shape investors’ incentives in

* Corresponding author. Universidade de Santiago de Compostela. Departamento de Fundamentos del An
alisis Econ

omico. Avda. Xo
an XXIII, s/n, 15782 Santiago de
Compostela, Spain.
E-mail address: [email protected] (J.C. Reboredo).

https://doi.org/10.1016/j.econmod.2019.09.004
Received 9 May 2019; Received in revised form 14 August 2019; Accepted 5 September 2019
Available online xxxx
0264-9993/© 2019 Elsevier B.V. All rights reserved.

Please cite this article as: Reboredo, J.C., Ugolini, A., Price connectedness between green bond and financial markets, Economic Modelling, https://
doi.org/10.1016/j.econmod.2019.09.004
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

the mobilization of financial resources for green projects that, in turn, are model. Unlike the bivariate copula model that accounts for contempo-
essential to developing a low-carbon economy (OECD, 2016). raneous bivariate dependence (Reboredo, 2018), the VAR model accu-
This paper aims to study price transmission processes across the green rately characterizes multivariate dependence and the network links
bond market and other bond and financial markets whose conditions may across markets, capturing direct and indirect transmission of financial
shape the financial performance of green bonds; those markets include shocks across markets. The VAR approach has previously been used to
the treasury, corporate and high-yield corporate bond markets and the study credit risk spillovers among global financial institutions (Yang and
stock, currency and energy commodity markets. Although green bonds Zhou, 2013) and to examine volatility spillovers across different inter-
are priced using similar pricing factors as used for any other financial national financial markets (Yang and Zhou, 2017). Instead of imposing a
asset, the fact that green bond proceeds specifically finance environ- priori restrictions on the contemporaneous causality effects among
mentally beneficial projects raises the question as to how price oscilla- different markets, we identify structural VAR model parameters using the
tions in financial markets might affect the non-pecuniary motives that heteroskedasticity embedded in price changes in the assets traded in
particularly support the green bond price. Those non-pecuniary motives, different markets (Rigobon, 2003; Ehrmann et al., 2011). We do this in
tied to investors’ pro-environmental preferences, may buffer or exacer- such a way that we can test for the existence, size and significance of
bate the effect of price swings in financial markets on green bonds and contemporaneous effects among markets.
vice versa. We specifically address the degree of integration or frag- Second, from the forecast error variance decomposition of the struc-
mentation between the green bond market and several financial markets tural VAR model, we disentangle the portion of the forecast error vari-
by considering how contemporaneous and lagged price changes are ance of the green bond prices that can be explained by different shocks in
transmitted, so as to quantify the size of price return spillovers to/from the other markets, so as to measure price return spillovers from this in-
the green bond market and the other financial markets. formation, as proposed by Diebold and Yilmaz (2014). Our spillover
Previous empirical research on green finance has examined the per- measure is different, however, as it arises from a data-driven structural
formance of green financial stocks and the impact of price changes in VAR instead of from contemporaneous causality variable ordering as
other kinds of assets on green financial stocks. Ortas and Moneva (2013) given by a Cholesky-based variance decomposition or by an ordering-free
find that the return performance of clean-technology equity indices generalized variance decomposition as introduced by Pesaran and Shin
surpasses that of conventional stock indices but also displays greater risk. (1998). We also check for the statistical significance of our spillover
Several studies show that renewable energy firm prices are sensitive to measures using a Monte Carlo simulation approach.
oil price oscillations (Henriques and Sadorsky, 2008; Sadorsky, 2012a; Third, for the period October 2014 to June 2019 we report empirical
Managi and Okimoto, 2013; Kumar et al., 2012; Broadstock et al., 2012; evidence that indicates that the green bond market (represented by the
Inchauspe et al., 2015), that there are volatility spillovers between oil global Barclays MSCI Green Bond Index) is mainly integrated with the
and clean energy stock prices (Sadorsky, 2012b; Wen et al., 2014) and treasury bond market (represented by the Bloomberg Barclays Global
that there is causality and tail dependence between oil and renewable Treasury Index), which is consistent with evidence reported by Reboredo
energy firm prices (Reboredo, 2015; Reboredo et al., 2017a). Further (2018) but adds new information in that the green bond market receives
studies have analysed whether green mutual funds outperform conven- sizeable spillover impacts from price changes in the treasury market and
tional mutual funds (Climent and Soriano, 2011; Reboredo et al., 2017b). transmits negligible reverse effects. Likewise, we find that the green bond
As for the flourishing literature on green bonds, a number of studies market is closely linked to the USD currency market, with fluctuations in
have examined green bond market volatility (Pham, 2016) and the use- the USD transmitting considerable spillovers to the green bond market,
fulness of green bonds for funding climate change costs (Flaherty et al., and fluctuations in the green bond market conveying negligible spillovers
2016). Other studies document the benefits of issuing green bonds for to the USD exchange rate. We also confirm that there is two-way spillover
both investors and issuers. Tang and Zhang (2019) find that stock prices between the green bond market and the corporate debt market (repre-
positively respond to green bond issuance announcements and that stock sented by Barclays Global Aggregate Corporate Index) and that the green
liquidity and institutional ownership are improved after the issuance of bond market is uncoupled with the high-yield corporate debt market
green bonds. Similarly, Flammer (2018) shows that the issuance of green (represented by Barclays Global High-Yield Index). We also document
corporate bonds has positive effects on financial and environmental that the green bond market is weakly connected with the stock market
performance and increases green innovations and long-term green in- (represented by the MSCI World Index) and independent of the energy
vestments. Regarding green bond pricing, several studies have analysed market (represented by the S&P GSCI Energy Spot Index), as demon-
price differentials between green and conventional bonds. Zerbid (2019) strated by negligible price spillover effects. Overall, our results point to
finds that the yield for green bonds is, on average, two basis points lower the fact the small green bond market essentially closely mirrors the
that the yield for conventional matched bonds. Similarly, other authors evolution of the treasury market, is impacted by the evolution of the USD
(Bachelet et al., 2019; Kapraun and Scheins, 2019; Baker et al., 2018; exchange rate and exhibits low integration with the high-yield corporate
Hachenberg and Schiereck, 2018) also find that green bonds pay lower bond and energy markets. Accordingly, the green bond market is a net
returns than conventional bonds. Another strand of the green bond spillover receiver, with a weak transmission capacity for spillover effects,
literature examines the relationship between green bond markets and which is consistent with the fact that green bonds still represent a small
other markets. Specifically, Broadstock and Cheng (2019) analyse the fraction of trading assets in the bond and equity markets. Our findings on
nexus between green and black bond price benchmarks, finding that price spillovers between the green bond and financial markets provide
correlation is sensitive to financial market conditions such as volatility, new insights for green investors managing the impact of movements in
economic policy uncertainty and news-based sentiment regarding green financial markets on their green bond positions. Specifically, our evi-
bonds. Likewise, Febi et al. (2018) report that market liquidity impacts dence indicates that green bonds behave as a specific asset class linked to
the yield spread of green bonds. Using bivariate copula functions, treasury bonds and currency exchange rates but offer some diversifica-
Reboredo (2018) shows that green bond prices co-move with treasury tion opportunities regarding other kinds of assets. This could raise in-
and corporate markets and move independently of stock and energy price vestors’ interest in green bonds and so favour increased flows to
changes. low-carbon projects, giving issuers the opportunity to enlarge their
This paper, in addressing price transmission between the green bond long-term green investor base.
market and other related financial markets, contributes to the extant The remainder of the paper is laid out as follows. In Section 2 we
green bond literature in three ways. describe the main features of our data for the green bond market and
First, we uncover the contemporaneous causal relations and lagged global markets for treasury bonds, corporate bonds, high-yield corporate
interdependence between price changes in the green bond market and debt, stocks and energy commodities. In Section 3, in order to assess
other financial markets using a structural vector autoregression (VAR) connectedness between green bond and financial markets, we outline our

2
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

structural VAR approach and heteroskedasticity identification approach performance across 23 developed markets countries, covering approxi-
and describe variance decomposition. In Section 4 we discuss our mately 85% of the free float-adjusted market capitalization in those
empirical results and the robustness analyses. Finally, Section 5 sum- countries. Finally, the energy commodity market is represented by the
marizes our results and concludes the paper. S&P GSCI Energy Spot CME Index (Energy), which is a production-
weighted index that reports reliable information to investors on the
2. Data performance of energy commodities.
Data for the green bond and financial market indices were sourced
Nowadays, the green bond market is a global market with large and from Bloomberg, with the exception of TWEXB data, which was sourced
diverse set of issuers — large companies, public entities and suprana- from the Federal Reserve Bank of St. Louis. The data refer to the period
tional institutions — and investors around the globe. To account for the 14 October 2014 to 25 June 2019, with the starting date determined by
financial performance of the increasingly large universe of green bonds, data availability for the GBI. The number of daily observations totalled
different global green bond indices have been created, including the 1227. The temporal dynamics of the financial market indices, along with
Barclays MSCI Green Bond Index, the S&P Dow Jones Green Bond Index; the GBI, is depicted in Fig. 1. As would be expected, the global green
the Solactive Green Bond Index and the Bank of America Merrill Lynch bond market closely co-moves with the Treasury and Corporate markets,
Green Bond Index. Each index reports on the performance of green bonds given that green bonds included in the GBI are mainly issued by gov-
and uses its own methodology and criteria for including bonds in the ernments and corporations; thus, the rates green bonds pay should be in
components of their index. As all of these indices exhibit similar dy- line with those paid by other fixed-income instruments that are issued by
namics and show a near-one correlation coefficient (see Reboredo, governments and corporations. As for High-Yield, Fig. 1 shows positive
2018), in this research we develop our analysis by considering the Bar- dependence between that market and the green bond market, even
clays MSCI Green Bond Index (GBI) as representative of the global green though both markets decouple in specific time periods — a fact that is
bond market.1 GBI, launched in November 2014, includes investment consistent with the idea that green and high-risk bonds are dissimilar
grade-rated green bonds with fixed-rate coupons issued in different financial instruments. As for the currency market, Fig. 1 shows that green
currencies by corporations and government/government-related and bonds and the USD rate move in opposite directions; thus, when the USD
securitized bonds. These are categorized as green by MSCI ESG Research depreciates (TWEXB goes up), the GBI moves down; this is consistent
(covering alternative energies, energy efficiency, pollution prevention with the fact that green bonds are issued in different countries in local
and control, sustainable water, green building and climate adaption), currencies but are expressed in USD in the GBI. For the MSCI World and
which uses rules that are consistent with the GBP. The index is reba- Energy indices, Fig. 1 shows that those indices weakly co-move with the
lanced every month and bonds are held in the index until final maturity. GBI, displaying clear graphical evidence of no dependence.
As the green bond market is globally represented, to examine spill- Table 1 shows descriptive statistics for the first difference of the
overs from/to the green bond to/from financial markets, we take infor- natural log of the GBI and financial market indices. For all series, average
mation about other related financial markets at the global level. Those daily values are close to zero and standard deviations are greater than
markets include: (a) fixed-income markets, including the treasury, average values in an order of several magnitudes. Moreover, standard
corporate debt and high-yield corporate debt markets, given that fixed- deviations confirm that green bonds have a volatility level between
income instruments are financially related to green bonds; (b) currency Treasury and Corporate bonds, and are less volatile than TWEXB, MSCI
and stock markets, as the dynamics of both stocks and the USD are related World and Energy. All series exhibit negative skewness and fat-tail dis-
to the risk-return profile of green bonds; and finally, (c) energy com- tributions and, consistently, the Jarque-Bera (JB) test rejects the
modity markets, as the evolution of energy prices influences the viability normality of the unconditional distribution. The Ljung-Box (LB) statistic
of green projects (Reboredo, 2015) funded by means of resources ob- suggests the absence of serial correlation in the return series (except for
tained from green bonds. High-Yield and MSCI World). For the squared series, this statistic sug-
The above-listed financial markets are represented as follows. The gests the presence of heteroskedasticity — evidence also confirmed by
treasury market is represented by the Bloomberg Barclays Global Trea- the autoregressive conditional heteroskedasticity-Lagrange multiplier
sury Total Return Index Value (Treasury), reflecting the dynamics of (ARCH-LM) statistic. In addition, the Dickey and Fuller (1979) test for the
global government bond markets. This index tracks fixed-rate, local null of non-stationarity and the Kwiatkowski et al. (1992) test for sta-
graded currency government debt issued in 37 countries and denomi- tionarity confirm that all return series are stationary. Finally, the Pearson
nated in 24 unhedged currencies. The corporate debt market is repre- correlation coefficient indicates that GBI is highly and positively corre-
sented by the Barclays Global Aggregate Corporate Index (Corporate), lated with Treasury and Corporate, but exhibits high negative linear
which reflects global graded fixed-rate corporate debt from developed dependence with TWEXB. Likewise, GBI is positively correlated with
and emerging markets issuers, including unhedged multicurrency bonds High-Yield and displays near-zero linear dependence with MSCI World
within the industrial, utility and financial sectors. The high-yield and Energy. In addition, fixed-income markets are linearly dependent
corporate debt market is represented by the Bloomberg Barclays Global and there is negative dependence between Treasury and Corporate with
High-Yield Index (High-Yield), which is a multicurrency index that ac- TWEXB, MSCI World and Energy.
counts for the evolution of the global high-yield debt market as reflected
by the union of the US High-Yield, the Pan-European High-Yield and 3. Empirical methodology
Emerging Markets Hard Currency High-Yield Indices. The dynamics of
the USD currency market is represented by the Trade Weighted US Dollar Below we outline the empirical approach used to characterize the
Index (TWEXB), which is a weighted average of the foreign exchange contemporaneous and lagged dynamic interaction between the green
values of the USD against the currencies of a broad group of major US bond market and financial markets and to describe spillovers or
trading partners. The global stock market is represented by the MSCI connectedness effects between those markets.
World Index (MSCI World), which accounts for large and mid-cap equity

3.1. A structural VAR model for price dynamics

1
We choose this index as representative of the green bond market since it is
computed by Barclays, which also computes, using a similar methodology, the As the price dynamics in the green bond and financial markets may
other global market indices we use in our analysis. Empirical results from exhibit contemporaneous and lagged feedback effects, we adopt an
different green bond market indices (available on request) are similar to the empirical set-up for the dynamics of price changes in those markets as
results reported in this article. given by the following structural VAR model:

3
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Fig. 1. Time series plots for daily green bond price index and treasury, corporate, high yield, exchange, stock and energy markets for the period October 2014 to June
2019. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

A yt ¼ μ þ ΦðLÞyt1 þ εt ; (1) raneous spillovers, the α-parameters of the matrix A can be identified
using information from those residuals. However, the variance-
where yt is a (7  1) vector of endogenous variables that includes in- covariance matrix of the reduced-form residuals, Varðηt Þ ¼ Ω ¼
formation on log price changes for green bonds, treasury bonds, corpo- ðA1 Þ Σ ðA1 Þ’, is non-diagonal — given that matrix A has non-zero off-
rate debt, high-yield corporate debt, currency exchange rates, stock and diagonal elements — and provides only 28 moments, from which we
energy price indices. A is a (7  7) matrix, with ones along the main need to identify the 42 parameters included in matrix A and the seven
diagonal, whose off-diagonal elements αij , for i 6¼ j ¼ 1; :::; 7 capture variance parameters contained in matrix Σ. Thus, identification of the
contemporaneous spillovers from price changes in market j to market i. parameters of the structural model in Eq. (1) using the reduced-form VAR
ΦðLÞ is a (7  7) polynomial matrix that accounts for the impact of the model is not feasible.
lagged value of endogenous variables until lag p on yt . Finally, μ is a Since the structural parameters in matrix A cannot be directly iden-
(7  1) vector of constants and εt is a (7  1) vector of unobserved tified from the reduced-form VAR, different practical identification
structural uncorrelated innovations with a diagonal variance-covariance strategies have been proposed in the literature (see, e.g., Mallick and
matrix Σ. Sousa, 2013; Granville and Mallick, 2009), such as (a) using a Choleski
The coefficients of matrix A, which cannot be recovered from the decomposition that imposes a recursive contemporaneous causal struc-
structural model using standard techniques, can be obtained by esti- ture; (b) imposing coefficient and/or sign restrictions on the parameters
mating the reduced-form VAR model: of matrix A; and (c) imposing theoretical restrictions on the value of the
long-run parameters. Any of those assumptions could be quite restrictive
yt ¼ ω þ BðLÞyt1 þ ηt ; (2)
and compelling in our analysis, because there are feedback effects among
green and financial markets that must be identified without imposing any
where ω ¼ A1 μ, BðLÞ ¼ A1 ΦðLÞ, and where the reduced-form re-
prior restrictive information regarding co-movement. We therefore need
siduals, ηt ¼ A1 εt , are related to the structural model residuals through
to be agnostic about the size and sign of the contemporaneous effects
matrix A. As reduced-form residuals embed information on contempo-
reflected in matrix A. For identification purposes, therefore, we use the

4
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Table 1
Descriptive statistics.
GBI Treasury Corporate High Yield TWEXB MSCI World Energy

Mean 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Std. Dev. 0.003 0.004 0.002 0.002 0.004 0.007 0.020
Minimum 0.016 0.021 0.011 0.014 0.024 0.050 0.093
Maximum 0.016 0.019 0.011 0.010 0.025 0.030 0.089
Skewness 0.157 0.063 0.367 0.464 0.114 0.711 0.032
Kurtosis 4.604 5.597 4.596 7.245 5.972 7.418 5.086
JB 136.275* 345.154* 157.479* 963.614* 453.411* 1099.516* 222.375*
LB(25) 27.50 29.49 32.27 163.36 19.31 60.50 27.45
[0.33] [0.24] [0.15] [0.00] [0.78] [0.00] [0.33]
LB(25)(2) 96.14 216.69 87.96 201.32 110.49 248.46 273.02
[0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
ARCH-LM 63.88 134.52 59.02 116.91 71.42 123.61 105.62
[0.00] [0.00] [0.00] [0.00] [0.00] [0.00] [0.00]
ADF 24.36* 24.02* 25.08* 20.23* 24.06* 23.95* 25.13*
KPSS 0.24 0.15 0.28 0.16 0.24 0.04 0.24
Correlation matrix
GBI 1
Treasury 0.89 1
Corporate 0.88 0.84 1
High Yield 0.38 0.22 0.36 1
TWEXB 0.81 0.73 0.63 0.42 1
MSCI World 0.04 0.10 0.03 0.60 0.14 1
Energy 0.02 0.05 0.06 0.33 0.19 0.37 1

Notes: The table reports descriptive statistics for daily price returns for the green bond and financial market indices indicated in each column for the period 14 October
2014 to 25 June 2019. JB denotes the Jarque-Bera statistic for normality; an asterisk (*) indicates rejection of the null hypothesis at the 5% level. LB(25) and LB(25)(2)
denote the Ljung-Box statistics for serial correlation in the returns and squared returns series, respectively, computed using 25 lags, with p values reported in square
brackets. ARCH-LM denotes Engle’s LM test for conditional heteroskedasticity, computed using 20 lags, with p values reported in square brackets. ADF and KPSS are the
empirical statistics of the Augmented Dickey and Fuller (1979) unit root test and the Kwiatkowski et al. (1992) stationarity test, respectively, with an asterisk (*)
indicating rejection of the null hypothesis at the 5% level.

heroskedasticity approach as developed by Sentana and Fiorentini 3.2. Spillover measures

(2001) and Rigobon (2003), which exploits the fact that the price return
series are heteroskedastic. Specifically, as in Ehrmann et al. (2011), we Once the structural model has been identified, the dynamic structure
assume that there are several heteroskedastic regimes with different and the economic interpretation of green bond-financial market inter-
variance-covariance matrices for the shocks and also assume that matrix dependence can be summarized by rewriting the reduced-form VAR
A remains stable across regimes. Thus, the existence of different volatility model in Eq. (2) as an infinite moving average (MA) process:
regimes adds more equations than unknowns and the identification
X
∞
problem can be solved using information from the residuals of the yt ¼ ξ þ Ψ j ηtj ; (3)
reduced-form model when at least two regimes can be found. j¼0
The identification strategy we apply involves several steps. First, we
estimate the reduced-form VAR model in Eq. (2) with a number of p lags where the MA coefficients are Ψ j ¼ B1 Ψ j1 þ B2 Ψ j2 þ :::: þ Bp Ψ jp , with
P
that removes serial correlation, so that the estimated residuals have a Ψ 0 as the identity matrix, with Ψ j ¼ 0 for j < 0, and where ξ ¼ ∞ j¼0 Ψ j ω.
similar contemporaneous dependence to the residuals of the structural Since the MA coefficients in Eq. (3) contain information on the accu-
model. Second, we compute the time-varying volatility for the residual mulated effects of a specific shock over a future horizon, summing all
series of each variable using a realized volatility measure (see McAleer those coefficient matrices we have the long-run effects or total multi-
and Medeiros, 2008), computed as the sum of the daily squared returns pliers. Likewise, the h-step-ahead forecast error variance decomposition
for a 40-day (approximately two trading months) rolling window over can be obtained from the matrix Ψ j . That is, the error from the h-step-
the whole sample period. The computed realized volatility at each time
ahead forecast of yt conditional on information a t is given by et;h ¼
and for each series is then assigned to a high-volatility regime when this Ph1
realized volatility is greater than its mean plus one standard deviation; j¼0 Ψ j ηtþhj and its variance-covariance matrix is given by:

otherwise it is applied to a low-volatility regime. Third, we restrict the

X
h1
whole set of regimes to seven (accounting for the seven markets exhib- Covðet;h Þ ¼ Ψ j Ω Ψ ’j (4)
iting high volatility) plus an additional tranquil regime where all markets j¼0
are low-volatility regimes. Fourth, we estimate the parameters of matrix
Thus, from the forecast error variance we can separate which portion
A using the generalized method of moments to minimize the following
of the h-step-ahead forecast error variance for each variable in yt can be
distance: ming’g with g ¼ ðA1 Þ Σ j ðA1 Þ’  Ωj for j ¼ 1, …,8 regimes.
explained by different shocks in the system. The contribution of each
Finally, to assess the significance of the α-parameters of matrix A, we use
variable is determined, among other things, by the α-parameters of ma-
a block-bootstrap procedure that functions as follows: (a) for each
trix A (note that Ω ¼ ðA1 Þ Σ ðA1 Þ’), which reports information on the
regime, we simulate pseudo-residuals with the same covariance structure
contemporaneous causal structure among variables that is obtained from
as the residuals; (b) from the simulated residuals and the estimated co-
the data.
efficients of the reduced VAR model, we compute pseudo-data y~t and
In order to compute the contribution of innovations in one variable to
obtain new estimated residuals for each regime from which we
the forecast error variance of other variables, we can rewrite Eq. (4) by
re-estimate a new matrix A; and, finally, (c) we repeat the previous steps
considering a (7  7) matrix C such that Σ ¼ CC’. Hence, since Ω ¼
1000 times and from these iterations we compute the distribution for the P ’
α-parameters of matrix A. A1 CC’ ðA1 Þ’, Eq. (4) can be written as Covðet;h Þ ¼ h1
j¼0 Θj Θj , where
Θj ¼ Ψ j A1 C. We can therefore express the contribution of the

5
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

innovations in market z to the h-step-ahead forecast error variance of Finally, for each θhk←z we build the 95% confidence intervals using a
market k as: Monte Carlo simulation of the estimated reduced-form VAR model in Eq.
(2) with 1000 replications. Furthermore, using Monte Carlo simulations,
hP
1
2
ða’k Θj az Þ we test whether a specific market k, i.e., the green bond market, is a net
j¼0 receiver of shocks with respect to another market z (or the set of other
θhk←z ¼ ; (5)
P
h1 P
7
markets) by formulating, respectively, the following two null hypotheses:
ðΘkz;j Þ2
j¼0 z¼1
h h
H0 : b
θ k←z > b
θ z←k ; (6)
where az is the zth column of the identity matrix I7 and where Θkz;i is the
kzth element of the matrix Θj . Therefore, the parameter θhk←z reports in- X
7 X
7
h h
formation on the proportion of the h-step-ahead forecast error of variable H0 : b
θ k←z > b
θ z←k (7)
k that can be explained by shocks in variable z. z¼1 k¼1
k6¼z k6¼z
As in Diebold and Yilmaz (2012, 2014), we use information contained
in the parameters θhk←z , for k; z ¼ 1; :::; 7 to measure spillover effects from
4. Empirical results
market z to market k, and so obtain a (7  7) connectedness matrix where
the main diagonal accounts for the contribution of market z to its own
We estimate the structural VAR for the joint price dynamics between
forecast error variance and the off-diagonal elements account for the
green bonds and financial markets, reporting results, focusing first on
impact of shocks between market z and market k. Thus, each row of this
contemporaneous price spillovers, then on price connectedness measures
matrix accounts for the contribution of the different markets, indicated in
and finally presenting a robustness check.
columns, on the forecast error variance of the market indicated in the
row, whereas each column of this matrix reports information on the
4.1. Co-movement between green bond and financial markets
impact that the market indicated in the column conveys to the different
markets indicated in rows. Our spillover measure differs from the Die-
Table 2 reports estimates for overall and direct price spillovers as
bold and Yilmaz (2014) approach in that we obtain the variance
decomposition from a structural VAR model; thus, the contemporaneous given by matrices A1 and A, respectively. These matrices are computed
causality among dependent variables is obtained from specific (hetero- from the estimated residuals of the reduced-form VAR model in Eq. (2),
skedastic) data features instead of by imposing arbitrary contempora- with two lags selected according to Schwarz’s Bayesian criterion. Each
neous causality (such as that given by the Choleski-based variance row of the matrix reports the estimated impacts that one financial market
decomposition or that based on the Pesaran and Shin (1998) —indicated in that row — receives from markets indicated in the col-
ordering-free generalized variance decomposition). umns, whereas the parameters reported in each column account for the
estimated impacts that the market indicated in the column transmit to

Table 2
Parameter estimates of the structural parameters for matrix A.
Panel A. Estimates for matrix A1 .

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 1.496* 0.518* 0.523* 0.335* 0.666* 0.111* 0.005

(0.112) (0.050) (0.103) (0.042) (0.016) (0.012) (0.005)
Treasury 0.773* 1.498* 0.506* 0.167 0.709* 0.067* 0.012*
(0.242) (0.074) (0.192) (0.097) (0.027) (0.013) (0.006)
Corporate 0.470* 0.338* 1.235* 0.252* 0.296* 0.043* 0.013*
(0.090) (0.033) (0.068) (0.017) (0.012) (0.008) (0.004)
High Yield 0.228 0.261* 0.209 1.324* 0.100* 0.173* 0.029*
(0.276) (0.037) (0.222) (0.070) (0.018) (0.005) (0.004)
TWEXB 0.679 0.504* 0.534* 0.477* 1.423* 0.181* 0.029*
(0.664) (0.071) (0.128) (0.069) (0.024) (0.014) (0.006)
MSCI World 0.101 0.866* 0.117 1.067* 0.213* 1.115* 0.109*
(0.559) (0.102) (0.464) (0.165) (0.069) (0.018) (0.011)
Energy 0.210 0.377* 0.050 0.716* 0.166* 0.246* 1.032*
(0.241) (0.135) (0.213) (0.085) (0.055) (0.036) (0.005)

Panel B. Estimates for matrix A.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 1 0.228* 0.174 0.136* 0.312* 0.010 0.014*

(0.028) (0.172) (0.032) (0.042) (0.010) (0.004)
Treasury 0.358 1 0.182 0.387* 0.316* 0.026 0.008
(0.191) (0.139) (0.066) (0.064) (0.022) (0.006)
Corporate 0.190 0.170* 1 0.178* 0.020 0.018* 0.013*
(0.141) (0.027) (0.031) (0.057) (0.008) (0.004)
High Yield 0.190 0.225* 0.159 1 0.078* 0.128* 0.013*
(0.185) (0.096) (0.131) (0.034) (0.016) (0.004)
TWEXB 0.197 0.318* 0.183 0.245* 1 0.074* 0.021*
(0.214) (0.056) (0.119) (0.067) (0.037) (0.006)
MSCI World 0.064 0.442* 0.061 0.704* 0.005 1 0.080*
(0.201) (0.188) (0.244) (0.305) (0.182) (0.029)
Energy 0.154 0.209 0.046 0.309 0.158 0.146 1
(0.167) (0.179) (0.183) (0.247) (0.259) (0.185)

Notes: The table reports estimates of the parameters of the matrices A1 and A for the reduced-from model in Eq. (2) using identification through heteroskedasticity
approach. Markets that receive shocks are indicated in rows, while columns indicate markets that spawn shocks. Asterisks denote statistical significance at the 95% level,
obtained using a bootstrap approach with 1000 replications.

6
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

markets indicated in the rows. Using a bootstrap procedure with 1000 parameter estimates, as depicted in Figs. 2 and 3.
replications, we obtain the standard errors of the parameters (reported in Regarding the overall effects between green bond and financial
parenthesis), the statistical significance (indicated with asterisks) of the markets, the first row of matrix A1 in Panel A of Table 2 shows that GBI
estimated parameters according to p-values, and the distribution of the changes in response to price shocks originating in all markets (except

Fig. 2. Distributions of the coefficients of matrix A1 using 1000 bootstrap replications.

Fig. 3. Distributions of the coefficients of matrix A using 1000 bootstrap replications.

7
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Table 3
Estimates of the market connectedness matrix from the reduced-form model.
GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 31.51 29.43 12.74 2.77 21.98 1.40 0.17

(14.68; 40.41) (23.98; 62.66) (5.31; 15.01) (1.02; 3.97) (9.10; 32.52) (0.88; 2.44) (0.01; 1.81)
Treasury 2.73 75.31 4.65 0.18 16.02 0.52 0.58
(0.71; 5.31) (68.93; 92.01) (1.07; 8.11) (0.03; 1.05) (4.09; 20.19) (0.21; 1.34) (0.07; 2.13)
Corporate 4.91 27.03 53.72 3.25 7.76 0.73 2.60
(2.25; 8.01) (23.22; 60.88) (27.03; 57.04) (1.31; 4.71) (3.24; 13.44) (0.27; 2.26) (0.31; 6.18)
High Yield 2.55 0.61 6.03 59.22 1.95 11.39 18.24
(1.26; 5.73) (0.22; 2.97) (3.32; 9.41) (35.46; 72.56) (0.84; 4.11) (7.97; 21.16) (3.04; 28.98)
TWEXB 3.09 14.87 4.57 3.42 63.10 2.72 8.23
(1.38; 5.25) (12.13; 42.32) (1.87; 6.86) (1.36; 4.89) (36.97; 73.78) (1.55; 4.42) (1.17; 12.87)
MSCI World 0.37 3.97 2.02 8.45 0.12 68.48 16.58
(0.19; 1.67) (2.83; 11.49) (0.82; 4.10) (3.24; 12.38) (0.08; 0.89) (59.41; 81.96) (2.96; 20.52)
Energy 0.52 0.05 0.06 0.15 0.01 0.32 98.89
(0.12; 6.54) (0.03; 2.08) (0.03; 2.29) (0.05; 2.35) (0.01; 1.04) (0.24; 2.90) (87.27; 99.15)

Notes: The table reports estimates of the connectedness matrix computed according to Eq. (5) and using a forecast horizon of h ¼ 10 trading days. Each row accounts for
the percentage fraction of the forecast error variance of the market indicated in the row that is explained by the markets indicated in columns. Values in each row sum to
100%. Each column reports the contribution of the variance of the market indicated in the column to each market indicated in rows. Reported in brackets are 95%
confidence intervals computed using 1000 Monte Carlo simulations of the reduced-form model in Eq. (2).

energy), although the size of those effects varies. Thus, the sizes of the As for direct effects, obtained from matrix A as reported in Panel B of
reaction coefficients indicate that the reaction of green bonds to shocks in Table 2, the estimated coefficients included in the first row indicate that
the treasury, corporate and currency markets is more sizeable than the the green bond market receives statistically significant spillover effects
reaction to shocks arising from other markets. Specifically, the estimated from the treasury, high-yield and currency markets, in such a way that
coefficients imply that an increase (decrease) of 1% in treasury and upward (downward) movements in treasury or high-yield markets lead
corporate price returns produces an increase (decrease) in the green bond to upward (downward) movements in the green bond market, whereas a
price index of 0.51% and 0.52%, respectively; whereas a depreciation depreciation in the USD reduces the value of the green bond market.
(appreciation) in the USD of a similar size produces a decrease (increase) Likewise, empirical estimates indicate that green bonds receive a statis-
in green bond returns of 0.66%. This evidence shows that the green bond tically significant impact from energy markets, although those effects are
market closely co-moves with the treasury, corporate and currency small in size. In contrast, price effects from the stock market do not have a
markets. The explanation is that the green bond index includes a relevant statistically significant impact on green bond prices. The evidence on
proportion of green bonds issued by governments, supranational entities direct effects to the green bond market is consistent with the findings
and firms, who set yields for green bonds according to government and reported in the first row of matrix A1 ; there is, however, a considerable
corporate bond yields, which explains why both bond classes exhibit difference in the size of the reaction coefficients, which are lower for
some similar features; furthermore, the global green bond index includes direct spillovers than for overall spillovers.
bonds issued in different currencies but reflected in the green bond index The first column of matrix A shows that price changes in green bond
in USD. Likewise, our evidence indicates that green bonds co-move with markets have no significant impact on other financial markets. This is
high-yield bonds issued by corporations, although the intensity of co- consistent with the fact that the green bond market is still small in size, so
movement is considerably lower. Finally, our evidence indicates that green bond price oscillations generally go unnoticed in other financial
green bonds are positively impacted by shocks in stock markets and markets.
receive no significant impacts from energy returns. In considering other markets, the empirical evidence shows that
As for the effect of shocks from the green bond market to other several markets, namely, the treasury, corporate and currency markets,
financial markets, the first column of matrix A1 in Table 2 indicates that move in tandem with significant one-way spillovers. Similar evidence is
those shocks have negligible effects on other markets, with the exception found for the energy and currency markets, stock and currency markets,
of the treasury and corporate debt markets, which receive the effects of and energy and high-yield debt markets. In contrast, spillovers are two-
shocks in green bond prices. This evidence is consistent with the fact that way between the stock and high-yield corporate debt markets and be-
the green bond market is still small in size and underrepresented as a tween the treasury and currency markets.
financial market, so changes in green bond prices generally go unnoticed
in other markets, particularly in large markets such as the currency, stock 4.2. Connectedness between green bond and financial markets
and energy markets.
Beyond the green bond market, the evidence in Panel A of Table 2 Having estimated the contemporaneous spillover effects between
confirms that markets systematically respond to their own shocks, as green bonds and financial markets, we next measure the relevance of
indicated by the fact that the diagonal coefficients of matrix A1 are all each financial market for the green bond market and vice versa using the
significant and non-negligible in size. Moreover, some off-diagonal co- forecast error variance decomposition of the estimated reduced-form
efficients are statistically significant, reflecting that shocks from currency VAR model described in Eq. (2), for two lags and a time horizon h of
markets are transmitted to treasury, corporate and stock markets and that 10 days.2 Table 3 reports the 7  7 financial market connectedness ma-
shocks from stocks are conveyed to fixed-income and currency markets. trix, where each cell indicates the directional spillovers — expressed as
Also, the empirical results indicate that shocks from corporate markets the share of the total variance of the market indicated in the row — that
are transferred to treasury returns and that treasury shocks are trans- are explained by the variance of the market indicated in the column.
ferred to fixed-income and currency markets. Finally, shocks from energy
markets are transmitted to all markets.
2
Overall, our results point to the fact that the green bond market is There is no selection method to determine the appropriate length of the
impacted by shocks occurring in all financial markets excepting energy, forecast horizon. Following Diebold and Yilmaz’s analysis, we report results for
whereas it has a limited capacity to transmit its own shocks to other a 10-day horizon; notwithstanding, we check the sensitivity of our results to the
choice of forecasting horizon by reporting results (see appendix) for other ho-
financial markets.
rizons, such as 5, 20 and 30 trading days.

8
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Own-market effects are reported along the main diagonal. For the spill- green bond market is significantly greater than the reverse impact. For
over parameters included in each cell we also report the 95% confidence corporate debt, our empirical evidence also indicates that shocks from
interval computed using 1000 Monte Carlo simulations of the reduced- the green bond to the corporate market are small and of a smaller size
form model. Fig. 4 depicts the distribution function for each estimated than the reverse impact. Likewise, our evidence indicates that the green
spillover using 1000 Monte Carlo simulations of the reduced-form model. bond market is responsible for a small share of the variances in high-yield
Consistent with our results for contemporaneous spillovers, we find corporate, stock and energy markets; furthermore, considering the null
that the variance of the green bond market is mainly driven by shocks hypothesis in Eq. (6), the green bond market receives/transmits similar
arising from the treasury and currency markets, accounting for about size shocks from/to the high-yield, stock and energy markets.
30% of the total green bond variance (the green bond own variance ac- The market network connectedness analysis in Table 3 also reveals
counts for 31% of the total). In contrast with the relevance of contem- that the treasury market is an important market in explaining the vola-
poraneous spillovers between corporate and green debt markets, the tility of corporate, and, to a lesser extent, currency markets. Furthermore,
variance of the corporate debt market only explains a small fraction of volatility shocks in the currency markets amplify the volatility of the
the total variance (around 13%); this may be explained by the fact that treasury and corporate markets. Additionally, corporate volatility ex-
second- and successive-round effects from changes in the variance of plains a relevant share of the volatility of the treasury and currency
corporate debt are transmitted to green bonds by means of other markets markets, whereas energy markets have the capacity to explain volatility
receiving the impact of changes in corporate debt. Likewise, our evidence in stock and high-yield markets. Finally, we find that a relevant fraction
of connectedness indicates that the green bond market is weakly tied to of the volatility in all markets is explained by their own volatilities.
the stock market, which accounts for about 1% of the total variance of the Fig. 5 graphically depicts the spillover effects between green bond
green bond market, and it is also weakly connected with the energy and and financial markets reported in Table 3. The size of each node in the
high-yield markets. Overall, this evidence suggests that the green bond network relating the set of markets is determined by both the contribu-
market currently lacks autonomy, as changes are mainly shaped by the tion in terms of effects of each market on other markets (sum of the co-
dynamics of other markets, in particular, the treasury, corporate and efficients in each column excluding own-market effects) and of other
currency markets. This can be explained by the fact that many green markets on any particular market (sum of the coefficients in each row
bonds included in the GBI are issued by governments from different excluding own-market effects). The colour of the edges connecting two
countries and are expressed in USD; consequently, movements in trea- markets reflects the size of the effect contribution/reception to/from
sury and currency markets have a sizeable impact on the dynamics of the other markets. Thus, the green bond market is a net receiver of shocks, as
GBI. are the corporate and high-yield markets, whereas the treasury and
Regarding the contribution of green bond market variance to other currency markets play a leading role in the transmission of shocks. The
markets, our evidence (reported in the first column of Table 3) confirms figure also shows strong connections of the treasury, currency and
that the green bond market has a limited capacity to influence other corporate markets with the green bond market, and weaker connections
markets, as it accounts for but a small fraction of the total variances of for other markets. The figure also reflects the strong connection between
other markets. As would be expected, the green bond market has a poor the stock and high-yield markets, the effects from the treasury to
capacity to explain price volatility in the treasury and currencies markets. corporate markets and the effects of the currency markets on the treasury
Furthermore, testing of the null hypothesis in Eq. (6) clearly reveals that markets.
the impact of the variance from the treasury and currency markets to the Finally, we test whether the green bond market is a net receiver or

Fig. 4. Distributions of the network connectedness parameters using 1000 Monte Carlo simulations.

9
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

thresholds and considering volatility greater than the mean plus two
standard deviations (instead of one standard deviation). Our empirical
results are fully consistent with those reported in Tables 2 and 3 Finally,
we check whether the length of the window in which we compute
volatility has any impact on results. In considering rolling window
lengths of 20 and 60 days, however, we find similar evidence on
connectedness as reported in Table 3.

5. Conclusions

Green bonds are a relatively new sustainability-oriented fixed-income

financial instrument that have recently been gaining prominence to raise
funds for projects with environmental benefits that are consistent with a
climate-resilient economy. As investors start to scrutinize and price
environmental risks, understanding price transmission between green
bonds and other financial instruments is decisive in shaping both the risk-
return profile of green bonds and private incentives for the mobilization
of financial resources for the green projects that are essential to transition
to a low-carbon economy.
We examine the transmission of contemporaneous and lagged price
changes between the green bond and several financial markets (treasury,
Fig. 5. Green bond and financial market spillovers. (For interpretation of the
corporate and high-yield corporate bonds and stock, currency and energy
references to colour in this figure legend, the reader is referred to the Web
version of this article.) commodity markets) using a structural VAR model, with structural pa-
rameters identified through the heteroskedasticity that features in the
price changes of the assets under study. From the forecast error variance
transmitter of shocks by considering the null hypothesis in Eq. (7). Evi-
decomposition of the structural VAR model, we extricate the portion of
dence from the Monte Carlo simulations report a p-value of 0.99, so we
the forecast error variance for each variable that can be explained by
were unable to reject the null that the green bond market is a net receiver
shocks from other variables and, following Diebold and Yilmaz (2014),
of shocks.
we measure spillovers from this information. Our empirical evidence for
the period October 2014 to June 2019 indicates that the global green
4.3. Robustness check bond market is closely linked to the global treasury bond and USD cur-
rency markets, receiving sizeable price spillover impacts from those
We examine the robustness of our previous results in different ways. markets and transmitting negligible reverse effects. Although smaller in
First, we consider whether our evidence of spillovers from/to financial size, we also confirm the existence of two-way price spillovers between
markets to/from the green bond market change over the sample period. the global green bond and corporate debt markets. In contrast, the green
To account for time-varying effects, we run our reduced-form VAR esti- bond market is weakly connected with the high-yield corporate bond
mates using a rolling window of 220 trading days (one trading year)3 and market. Our evidence also indicates that the green bond market is weakly
compute, for each time window, the value of the spillovers as reported in connected with the stock market, nor does it receive or transmit price
the first row and column of Table 3. Fig. 6, which shows the time series spillovers from/to energy markets. All in all, our empirical evidence is
plots for those parameter spillovers, indicates that the dynamics of the consistent with the fact that the green bond market is still a small market,
parameter estimates is relatively smooth, with no abrupt changes in closely mirrors the evolution of the treasury market, is impacted by the
values. This graphical evidence is consistent with our previous evidence evolution of the USD and exhibits a low degree of integration with other
on price spillovers between the green bond market and financial markets, financial markets, in particular, with the energy market. As a result, the
as reported in Table 3. green bond market is a net price-spillover receiver with a weak capacity
Second, we split our sample into two subsamples – from October 2014 to transmit price spillover effects.
to January 2016, when a downward trend in the value of green bonds is Our empirical findings on price spillovers between green bond and
observed and from February 2016 to June 2019, when the green bond financial markets provide new insights in terms of portfolio design and
market shows abrupt swings in dipping and then peaking (see Fig. 1). risk management decisions for green investors around the world. Spe-
Table 4, which reports results from the reduced-form VAR estimates for cifically, investors holding green bond positions should bear in mind that
the two subsamples, shows that evidence on spillovers from/to green the value of those positions are immune to price oscillations in energy
bond market is qualitatively and quantitatively similar for both sub- markets but are impacted by changes in the value of treasury bonds and
sample periods, and that subsample results are consistent with the of the USD. Likewise, green bond price behaviour regarding other assets
empirical evidence reported in Table 3 for the whole sample period. offers investors the opportunity to use green bonds as a specific asset
Finally, we check the sensitivity of our contemporaneous spillover class for the purpose of hedging portfolio risks and minimizing downside
results to the definition of volatility. Specifically, we consider volatility as risks when holding investment positions in stocks, energy assets or high-
given by the standard deviation of returns (Ehrmann et al., 2011), yield corporate debt. In addition, investors can also benefit from holding
computed for a 40-day rolling window. The empirical evidence for green bond positions in that this greens their portfolios and boosts their
connectedness reported in Table 5 points to similar conclusions to those ESG scores. This is particularly relevant for investors and fund managers
drawn from Table 3, indicating that our evidence is not sensitive to the concerned with the exposure of their portfolio to long-run environmental
definition of volatility. We further check whether our results are sensitive and regulatory risks associated with the transition to a climate-resilient
to the definition of volatility regimes by changing the volatility economy. Our results also have implications for green bond issuers. As
green bonds are a sound financial device for investors, issuers can use this
financial vehicle to expand and diversify their investor base and enhance
3
We also estimate time-varying spillovers by using rolling windows for 150 their corporate social responsibility, even though green bond emissions
and 300 trading days, finding similar empirical evidence as reported here for a involve higher costs due to green certification.
rolling window of 220 trading days. Finally, our evidence on the potential benefits of green bonds also has

10
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Fig. 6. Time-varying spillovers between green bond and financial markets. (For interpretation of the references to colour in this figure legend, the reader is referred to
the Web version of this article.)

Table 4
Estimates of the market connectedness matrix for two subsample periods.
Panel A. October 2014 to January 2016.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 36.50 22.89 12.63 2.74 22.42 2.77 0.05

(12.79; 41.03) (18.51; 36.70) (9.06; 18.93) (1.88; 4.53) (17.14; 36.89) (1.37; 5.93) (0.01; 1.04)
Treasury 1.10 69.70 5.26 0.02 21.57 2.26 0.08
(0.57; 6.35) (60.10; 72.92) (1.29; 13.88) (0.01; 1.50) (17.39; 24.82) (0.67; 6.00) (0.08; 0.88)
Corporate 2.45 29.90 51.57 3.35 11.66 0.45 0.61
(1.78; 8.58) (21.25; 36.02) (43.72; 59.15) (2.39; 5.63) (8.16; 14.86) (0.26; 2.31) (0.33; 2.21)
High Yield 2.60 0.81 5.32 68.75 3.48 12.94 6.09
(1.46; 9.25) (0.04; 3.73) (4.26; 7.45) (57.02; 73.59) (2.33; 5.22) (8.80; 19.05) (3.91; 11.26)
TWEXB 1.36 12.47 2.91 2.66 75.41 3.69 1.50
(1.09; 5.56) (9.91; 15.58) (2.17; 6.47) (1.91; 4.67) (65.81; 77.95) (2.20; 7.61) (1.28; 3.04)
MSCI World 0.77 4.07 1.37 8.77 0.16 80.74 4.12
(0.19; 6.07) (2.81; 6.99) (0.82; 4.75) (5.48; 13.14) (0.14; 1.78) (70.33; 84.58) (3.25; 7.68)
Energy 0.24 0.14 0.23 0.29 0.11 1.61 97.38
(0.02; 6.14) (0.09; 4.83) (0.09; 6.06) (0.21; 2.81) (0.02; 2.81) (0.79; 5.42) (84.05; 97.63)

(continued on next page)

11
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Table 4 (continued )
Panel B. February 2016 to June 2019.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

Panel B. February 2016 to June 2019.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 32.71 30.83 11.44 2.54 21.27 1.18 0.02

(8.79; 46.24) (21.76; 54.36) (3.15; 14.96) (0.43; 3.83) (3.69; 34.97) (0.65; 3.79) (0.00; 0.98)
Treasury 2.63 78.59 3.31 0.22 14.92 0.24 0.08
(0.40; 5.99) (68.32; 96.45) (0.47; 6.74) (0.01; 1.25) (1.47; 20.74) (0.11; 1.17) (0.03; 1.35)
Corporate 5.35 30.75 51.87 2.85 7.96 0.86 0.36
(1.58; 8.02) (21.87; 74.41) (15.71; 60.24) (0.49; 4.49) (1.31; 15.37) (0.23; 3.39) (0.19; 4.19)
High Yield 3.00 0.28 8.82 69.15 2.49 13.86 2.41
(1.11; 5.27) (0.00; 2.10) (3.55; 12.64) (23.48; 77.69) (0.60; 5.66) (8.85; 35.65) (1.45; 28.39)
TWEXB 3.43 17.70 5.30 3.39 66.65 2.26 1.26
(0.77; 7.63) (11.49; 60.36) (1.18; 9.83) (0.74; 4.84) (17.77; 75.80) (1.05; 7.67) (0.64; 15.87)
MSCI World 0.41 5.70 4.03 9.78 0.19 76.78 3.10
(0.15; 1.19) (1.92; 15.56) (0.83; 7.11) (1.40; 15.10) (0.02; 1.05) (61.52; 84.52) (2.33; 17.21)
Energy 0.52 0.27 0.18 0.54 0.47 1.77 96.25
(0.01; 2.88) (0.03; 2.37) (0.02; 1.78) (0.02; 2.22) (0.01; 3.06) (0.28; 3.55) (89.55; 99.38)

Notes: See notes in Table 3.

Table 5
Estimates of the market connectedness matrix using standard deviations to account for volatility.
GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 40.99 18.34 6.88 3.35 16.14 0.31 3.99

(35.10; 46.10) (7.47; 9.13) (5.89; 7.53) (2.96; 3.74) (31.57; 39.62) (0.27; 0.34) (3.52; 4.44)
Treasury 7.06 41.66 7.50 1.81 29.36 10.32 2.29
(6.35; 7.73) (37.31; 46.48) (6.48; 8.32) (1.55; 2.02) (25.68; 32.97) (9.28; 11.39) (2.06; 2.57)
Corporate 5.05 11.13 49.16 0.28 18.81 5.94 9.64
(4.47; 5.57) (9.70; 12.55) (41.97; 55.26) (0.25; 0.32) (16.45; 20.96) (5.06; 6.64) (8.21; 10.49)
High Yield 2.38 0.71 1.24 29.46 9.66 54.69 1.86
(2.04; 2.59) (0.63; 0.78) (1.07; 1.36) (25.64; 32.56) (8.34; 10.83) (46.83; 60.82) (1.63; 2.06)
TWEXB 4.62 4.01 1.86 0.20 86.46 2.09 0.76
(4.15; 5.21) (3.53; 4.38) (1.65; 2.09) (0.18; 0.23) (73.63; 95.30) (1.81; 2.29) (0.68; 0.85)
MSCI World 0.47 1.00 0.31 0.52 0.18 93.89 3.63
(0.40; 0.53) (0.85; 1.09) (0.28; 0.35) (0.47; 0.59) (0.16; 0.20) (80.04; 102.11) (3.26; 4.03)
Energy 1.00 0.22 0.18 0.40 0.30 19.33 78.57
(0.87; 1.12) (0.19; 0.24) (0.15; 0.20) (0.35; 0.44) (0.26; 0.33) (16.56; 21.07) (69.23; 85.62)

Notes: See notes of Table 3.

policy implications for fiscal and monetary policies aimed at greening the Acknowledgements
economy. Specifically, in considering green fiscal policies or green sov-
ereign bonds to finance low-carbon transition policies, green bonds offer We thank the Editor (Sushanta Mallick), the associate editor and two
a solution with good acceptability in credit markets by investors, have anonymous reviewers for constructive comments and suggestions that
comprehensive effects on green growth and redistribute the cost of significantly improved the quality of the paper. We acknowledge finan-
climate change across generations; in contrast, green fiscal measures cial support from Agencia Estatal de Investigaci

on (Ministero de Ciencia,
affect household purchasing power in the short term. Some researches Innovaci
on y Universidades) under research project with reference
(e.g., Dafermos et al., 2018) have recently advocated the use of green RTI2018-100702-B-I00, co-funded by the European Regional Develop-
corporate bonds in a green quantitative easing programme aimed at ment Fund (ERDF/FEDER). Juan C. Reboredo acknowledges financial
reducing global warming and ameliorating climate-induced financial support provided by the Xunta de Galicia through research project
instability; such a programme would likely be more effective if green CONSOLIDACION
2019 GRC GI-2060 An
alise Econ
omica dos Mercados e
bonds could be used as a useful device for risk management purposes as ons – AEMI (ED431C 2019/11). Andrea Ugolini acknowledges
Instituci

implied by our results. financial support provide by the Brazilian National Council for Scientific
and Technological Development (CNPq).

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.econmod.2019.09.004.

Appendix

Tables A1-A3 shows similar results as Table 3, but considering different forecast horizons.

12
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

Table A.1
Estimates of the market connectedness matrix from the reduced-form model for a forecast horizon of h ¼ 5 trading days.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 31.51 29.43 12.74 2.77 21.98 1.40 0.17

(14.33; 43.79) (22.71; 60.43) (5.42; 15.41) (1.04; 4.22) (9.72; 32.61) (0.91; 2.48) (0.01; 1.78)
Treasury 2.73 75.31 4.65 0.18 16.02 0.52 0.58
(0.65; 5.60) (68.21; 91.38) (1.04; 8.14) (0.03; 1.08) (5.19; 20.95) (0.20; 1.41) (0.06; 2.22)
Corporate 4.91 27.03 53.72 3.25 7.76 0.73 2.60
(2.24; 8.14) (20.64; 59.96) (25.58; 59.30) (1.32; 4.93) (3.32; 13.78) (0.28; 2.11) (0.31; 5.56)
High Yield 2.55 0.61 6.03 59.23 1.95 11.39 18.24
(1.27; 5.66) (0.21; 2.85) (3.41; 9.47) (36.53; 72.73) (0.83; 4.13) (8.02; 20.43) (3.01; 28.15)
TWEXB 3.09 14.87 4.57 3.42 63.10 2.72 8.23
(1.26; 6.01) (11.27; 38.36) (1.80; 7.43) (1.36; 5.79) (38.46; 73.97) (1.65; 4.77) (1.15; 13.04)
MSCI World 0.37 3.97 2.02 8.45 0.12 68.48 16.58
(0.19; 1.63) (2.66; 11.34) (0.84; 4.49) (3.57; 12.49) (0.07; 0.91) (59.86; 82.10) (2.92; 20.37)
Energy 0.52 0.05 0.06 0.15 0.01 0.32 98.89
(0.13; 6.39) (0.03; 1.86) (0.03; 2.26) (0.06; 2.35) (0.01; 1.12) (0.24; 3.06) (87.83; 99.09)
Notes: This table replicates results reported in Table 3 but considering a forecast horizon period of 5 trading days.

Table A.2
Estimates of the market connectedness matrix from the reduced-form model for a forecast horizon of h ¼ 20 trading days.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 31.51 29.43 12.74 2.77 21.98 1.40 0.17

(14.99; 41.10) (22.06; 61.15) (5.64; 15.25) (1.10; 4.17) (9.25; 32.40) (0.89; 2.43) (0.01; 1.52)
Treasury 2.73 75.31 4.65 0.18 16.02 0.52 0.58
(0.72; 6.03) (67.26; 91.60) (1.10; 7.79) (0.03; 1.05) (4.59; 21.37) (0.23; 1.37) (0.06; 2.01)
Corporate 4.91 27.03 53.72 3.25 7.76 0.73 2.60
(2.29; 8.27) (21.72; 58.51) (28.00; 57.75) (1.43; 4.82) (3.34; 13.98) (0.29; 1.97) (0.29; 5.50)
High Yield 2.55 0.61 6.03 59.22 1.95 11.39 18.24
(1.28; 5.76) (0.20; 2.76) (3.26; 9.39) (36.32; 72.54) (0.99; 4.44) (7.82; 20.44) (2.86; 28.89)
TWEXB 3.09 14.87 4.57 3.42 63.10 2.72 8.23
(1.33; 5.35) (11.28; 37.81) (1.83; 7.33) (1.39; 4.93) (37.75; 73.95) (1.66; 4.43) (1.13; 12.17)
MSCI World 0.37 3.97 2.02 8.45 0.12 68.48 16.58
(0.20; 1.63) (2.80; 10.56) (0.81; 4.41) (3.58; 12.95) (0.09; 0.83) (58.57; 81.91) (2.86; 20.37)
Energy 0.52 0.05 0.06 0.15 0.01 0.32 98.89
(0.15; 6.12) (0.03; 1.87) (0.03; 2.24) (0.05; 2.48) (0.01; 1.03) (0.24; 3.03) (87.61; 99.07)
Notes: This table replicates results reported in Table 3 but considering a forecast horizon period of 20 trading days.

Table A.3
Estimates of the market connectedness matrix from the reduced-form model for a forecast horizon of h ¼ 30 trading days.

GBI Treasury Corporate High Yield TWEXB MSCI World Energy

GBI 31.51 29.43 12.74 2.77 21.98 1.40 0.17

(14.71; 42.42) (22.34; 60.04) (5.95; 15.45) (1.11; 4.27) (7.31; 33.18) (0.88; 2.38) (0.01; 1.61)
Treasury 2.73 75.31 4.65 0.18 16.02 0.52 0.58
(0.69; 5.38) (67.24; 91.71) (1.19; 8.23) (0.03; 1.04) (4.50; 22.42) (0.20; 1.39) (0.06; 2.08)
Corporate 4.91 27.03 53.72 3.25 7.76 0.73 2.60
(2.29; 8.31) (22.11; 57.31) (27.33; 57.73) (1.40; 4.99) (2.46; 14.30) (0.26; 2.03) (0.30; 5.25)
High Yield 2.55 0.61 6.03 59.22 1.95 11.39 18.24
(1.27; 5.73) (0.23; 3.01) (3.52; 9.33) (36.76; 72.98) (0.84; 4.18) (7.67; 20.32) (3.06; 28.54)
TWEXB 3.09 14.87 4.57 3.42 63.10 2.72 8.23
(1.33; 5.87) (11.13; 39.64) (1.82; 7.50) (1.41; 5.67) (33.40; 73.83) (1.63; 4.30) (1.17; 12.38)
MSCI World 0.37 3.97 2.02 8.45 0.12 68.48 16.58
(0.19; 1.69) (2.82; 11.50) (0.85; 4.32) (3.53; 13.15) (0.08; 0.86) (59.23; 82.30) (2.96; 19.90)
Energy 0.52 0.05 0.06 0.15 0.01 0.32 98.89
(0.15; 6.70) (0.04; 1.71) (0.03; 2.19) (0.06; 2.66) (0.01; 1.09) (0.24; 3.02) (87.56; 99.09)
Notes: This table replicates results reported in Table 3 but considering a forecast horizon period of 30 trading days.

References Climent, F., Soriano, P., 2011. Green and good? The investment performance of us
environmental mutual funds. J. Bus. Ethics 103 (2), 275–287.
Dafermos, Y., Nikolaidi, M., Galanis, G., 2018. Climate Change, financial stability and
Bachelet, M.J., Becchetti, L., Manfredonia, S., 2019. The green bonds premium puzzle: the
monetary policy. Ecol. Econ. 152, 219–234.
role of issuer characteristics and third-party verification. Sustainability 11 (4), 1098.
Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive time
Baker, M., Bergstresser, D., Serafeim, G., Wurgler, J., 2018. Financing the response to
series with a unit root. Journal of the American Statistical Society 75, 427–431.
climate change: the pricing and ownership of us green bonds. NBER Working Paper
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional
No, p. 25194.
measurement of volatility spillovers. Int. J. Forecast. 28, 57–66.
Broadstock, D.C., Cheng, L.T., 2019. Time-varying relation between black and green bond
Diebold, F.X., Yilmaz, K., 2014. On the network topology of variance decompositions:
price benchmarks: macroeconomic determinants for the first decade. Financ. Res.
measuring the connectedness of financial firms. J. Econom. 182, 119–134.
Lett. 29, 17–22.
Ehrmann, M., Fratscher, M., Rigobon, R., 2011. Stocks, bonds, money markets and
Broadstock, D.C., Cao, H., Zhang, D., 2012. Oil shocks and their impact on energy related
exchange rates: measuring international financial transmission. J. Appl. Econom. 26,
stocks in China. Energy Econ. 34, 1888–1895.
948–974.
Climate Bonds Initiative, 2018. Bonds and climate change: state of the market. https://
www.climatebonds.net/resources/reports/green-bonds-state-market-2018.

13
J.C. Reboredo, A. Ugolini Economic Modelling xxx (xxxx) xxx

European Commission, 2016. Study on the potential of green bond finance for resource- OECD, 2017. Mobilising bond markets for a low-carbon transition. https://www.oecd.or
efficient investments. http://ec.europa.eu/environment/enveco/pdf/potential-green g/env/mobilising-bond-markets-for-a-low-carbon-transition-9789264272323-en.h
-bond.pdf. tm.
Febi, W., Schafer, D., Stephan, A., Sun, C., 2018. The impact of liquidity risk on the yield Ortas, E., Moneva, J.M., 2013. The Clean Techs equity indexes at stake: risk and return
spread of green bonds. Financ. Res. Lett. 27, 53–59. dynamics analysis. Energy 57, 259–269.
Flaherty, M., Gevorkyan, A., Radpour, S., Semmler, W., 2016. Financing climate policies Pesaran, H.H., Shin, Y., 1998. Generalized impulse response analysis in linear
through climate bonds - a three stage model and empirics. Res. Int. Bus. Financ. 42, multivariate models. Econ. Lett. 58, 17–29.
468–479. Pham, L., 2016. Is it risky to go green? A volatility analysis of the green bond market.
Flammer, C., 2018. Corporate green bonds. Available at SSRN. https://papers.ssrn.com/s Journal of Sustainable Finance & Investment 6, 263–291.
ol3/papers.cfm?abstract_id¼3125518. Reboredo, J.C., 2015. Is there dependence and systemic risk between oil and renewable
Granville, B., Mallick, S., 2009. Monetary and financial stability in the Euro area: pro- energy stock prices? Energy Econ. 48, 32–45.
cyclicality versus trade-off. J. Int. Financ. Mark. Inst. Money 19 (4), 662–674. Reboredo, J.C., 2018. Green bond and financial markets: co-movement, diversification
Hachenberg, B., Schiereck, D., 2018. Are green bonds priced differently from and price spillover effects. Energy Econ. 74, 38–50.
conventional bonds? J. Asset Manag. 19 (6), 371–383. Reboredo, J.C., Rivera-Castro, M.A., Ugolini, A., 2017a. Wavelet-based test of co-
Henriques, I., Sadorsky, P., 2008. Oil prices and the stock prices of alternative energy movement and causality between oil and renewable energy stock prices. Energy
companies. Energy Econ. 30, 998–1010. Econ. 61, 241–252.
Inchauspe, J., Ripple, R.D., Trück, S., 2015. The dynamics of returns on renewable energy Reboredo, J.C., Quintela, M., Otero, L., 2017b. Do investors pay a premium for going
companies: a state-space approach. Energy Econ. 48, 325–335. green? Evidence from alternative energy mutual funds. Renew. Sustain. Energy Rev.
Kapraun, J., Scheins, C., 2019. (in)-credibly green: which bonds trade at a green bond 73, 512–520.
premium? Available at SSRN. https://ssrn.com/abstract¼3347337. Rigobon, R., 2003. Identification through heteroskedasticity. Rev. Econ. Stat. 85,
Kumar, S., Managi, S., Matsuda, A., 2012. Stock prices of clean energy firms, oil and 777–792.
carbon markets: a vector autoregressive analysis. Energy Econ. 34, 215–226. Sadorsky, P., 2012a. Modeling renewable energy company risk. Energy Policy 40, 39–48.
Kwiatkowski, D., Phillips, P.C.B., Schmidt, P., Shin, Y., 1992. Testing the null hypothesis Sadorsky, P., 2012b. Correlations and volatility spillovers between oil prices and the stock
of stationary against the alternative of a unit root. J. Econom. 54, 159–178. prices of clean energy and technology companies. Energy Econ. 34 (1), 248–255.
Mallick, S.K., Sousa, R.M., 2013. The real effects of financial stress in the Eurozone. Int. Sentana, E., Fiorentini, G., 2001. Identification, estimation and testing of conditionally
Rev. Financ. Anal. 30, 1–17. heteroskedastic factor models. J. Econom. 102, 143–164.
Managi, S., Okimoto, T., 2013. Does the price of oil interact with clean energy prices in Tang, D.Y., Zhang, Y., 2019. Do shareholders benefit from green bonds? J. Corp. Financ.
the stock market? Jpn. World Econ. 27, 1–9. forthcoming.
McAleer, M., Medeiros, M.C., 2008. Realized volatility: a review. Econom. Rev. 27, Wen, X., Guo, Y., Wei, Y., Huang, D., 2014. How do the stock prices of new energy and
10–45. fossil fuel companies correlate? Evidence from China. Energy Econ. 41, 63–75.
Monasterolo, I., Raberto, M., 2018. The EIRIN flow-of-funds behavioural model of green Yang, J., Zhou, Y., 2013. Credit risk spillover among financial institutions around global
fiscal policies and green sovereign bonds. Ecol. Econ. 144, 228–243. credit crisis: firm-level evidence. Manag. Sci. 59, 2343–2359.
OECD, 2016. Mobilising the Debt Capital Markets for a Low Carbon Transition. OECD Yang, Z., Zhou, Y., 2017. Quantitative easing and volatility spillovers across countries and
Publishing, Paris. asset classes. Manag. Sci. 63, 333–354.
Zerbid, O.D., 2019. The effect of pro-environmental preferences on bond prices: evidence
from green bonds. J. Bank. Financ. 98, 39–60.

14