Time and frequency dynamics of connectedness between renewable energy stocks and crude oil prices

Full text

1
Time and frequency dynamics of connectedness between renewable
energy stocks and crude oil prices
Román Ferrer
Department of Actuarial and Financial Economics, University of Valencia, Spain
Email: roman.ferrer@uv.es
Syed Jawad Hussain Shahzad
Energy and Sustainable Development (ESD), Montpellier Business School, France
Email: j.syed@montpellier-bs.com
Raquel López
Department of Economics and Finance, University of Castilla-La Mancha, Spain
Email: Raquel.Lopez@uclm.es
Francisco Jareño
Department of Economics and Finance, University of Castilla-La Mancha, Spain
Email: Francisco.Jareno@uclm.es
Abstract
This paper examines the time and frequency dynamics of connectedness among stock
prices of U.S. clean energy companies, crude oil prices and a number of key financial
variables using the methodology developed by Barunik and Krehlik (2018). This
approach allows measuring the dynamics of return and volatility connectedness over time
and across frequencies simultaneously. The empirical results show that most of return and
volatility connectedness is generated in the very short-term, i.e. movements up to five
days, while the long-term plays a minor role. Our analysis further reveals a greater degree
of interconnectedness across crude oil and financial markets since the onset of the U.S.
subprime mortgage crisis in summer of 2007, consistent with the view of a global re-
pricing of risk triggered by the recent worldwide financial crisis. Crude oil prices do not
appear as a key driver of the stock market performance of renewable energy companies
in the short-term or the long-term, which suggests a decoupling of the alternative energy
industry from the traditional energy market. Moreover, crude oil prices are a net receiver
of financial shocks, supporting the financialization of the commodity markets since the
early 2000s. In addition, a significant pairwise connectedness is found, mainly in the
short-term, between clean energy and technology stock prices, indicating that these two
types of stocks are perceived by investors as similar assets. These results can have
important practical implications for investors and policy makers with different time
horizons.
Keywords: connectedness, renewable energy stocks, crude oil price, information
transmission, time-frequency space.

2
1. Introduction
Renewable energy has gained considerable ground worldwide as a viable energy
alternative due to a combination of factors, such as growing international concern about
climate change, fossil fuel depletion, energy security issues, technology innovation, and
high and volatile prices of petroleum-based fuels.1 In recent years, the clean energy sector
has undergone record developments and trends. According to the Global Status Report
published by REN21 (2017), about 19.3% of global final energy consumption was
supplied by renewable energy in 2016.2 In its World Energy Outlook published in 2017,
the International Energy Agency (IEA) confirms that alternative sources of energy
currently cover about 40% of the increase in primary demand, so renewables reached new
records around the world in 2016 (IEA, 2017). In addition, the International Renewable
Energy Agency (IRENA) states that the renewable energy market increased the total
capacity of renewable energy by 8.8% in 2016 (IRENA, 2017), thus reaching the largest
global capacity enlargement never seen before.
Despite the tremendous development of the alternative energy sector over the past few
years, crude oil remains the largest source of primary energy, accounting for a third of
global energy consumption in 2016 (BP Statistical Review of World Energy 2017). The
strength of the renewable energy sector, together with the still predominance of crude oil
and its extremely volatile behavior from mid-2008, have spurred a strong interest among
practitioners and academics recently in knowing whether oil prices are a major driver of
the financial performance of green energy companies. A positive relationship between
stock prices of clean energy firms and crude oil prices is typically hypothesized (e.g.,
Henriques and Sadorsky, 2008; Kumar et al., 2012; and Sadorsky, 2012). This positive
linkage has its roots in that clean energy is commonly viewed as a substitute to the fossil
fuel energy. Since investors and consumers seek cheaper alternatives to fossil fuels, rising
oil prices should encourage a substitution effect away from petroleum-based energy
towards alternative energy sources as the green energy sector becomes comparatively
more competitive. Hence, the positive effect of higher crude oil prices on the alternative
energy industry should lead to a substantial improvement in the stock market performance
of new energy companies. Another plausible explanation for the positive oil price-clean
1 It is worth highlighting that the terms renewable energy, clean energy, alternative energy, sustainable
energy, green energy and new energy are used interchangeably in this paper.
2 REN21 refers to Renewable Energy Policy Network for the 21st Century. REN21 is the global renewable
energy policy multi-stakeholder network that connects a wide range of key actors.

3
energy association is related to the fact that until recently renewable energy has been far
more expensive than traditional fossil fuels. The high costs of building and installing
renewable energy systems have historically represented a formidable barrier for
alternative energy. Therefore, any significant oil price drop limited seriously the
attractiveness and economic viability of clean energy projects, leading to a sudden halt in
the development of sustainable energy, with the consequent detrimental effect on stock
prices of new energy firms.
In this backdrop, the primary aim of this paper is to analyze the dynamic interdependence
among stock prices of U.S. renewable energy firms, crude oil prices and a number of key
financial indicators (i.e., stock prices of technology and conventional energy firms, U.S.
Treasury bond yields, the U.S. default spread and the volatility of the U.S. stock and
government bond markets) in the time-frequency space. To this end, the time-frequency
connectedness methodology recently developed by Barunik and Krehlik (2018) is
applied. This framework can be thought as the time-frequency version of the spillover
index approach of Diebold and Yilmaz (2012). While the Diebold-Yilmaz model focuses
on the time domain only, the approach of Barunik and Krehlik (2018) allows one to assess
the magnitude and direction of spillovers over time and across frequencies
simultaneously. Thus, apart from including the time-varying information of the method
of Diebold and Yilmaz (2012), the Barunik-Krehlik framework decomposes aggregate
connectedness into different frequency domains, enabling to determine the specific
frequencies that most contribute to the connectedness of a system.
The main reason to think that connectedness between crude oil and financial markets may
vary across frequencies is based on the heterogeneity of multiple economic agents
interacting in these markets. More precisely, market participants operate at diverse time
horizons (represented by frequencies) ranging from seconds to several years primarily
because they have very different beliefs, objectives, preferences and institutional
constraints as well as distinct levels of information assimilation and risk tolerance.
Therefore, economic and financial shocks can propagate through markets producing
heterogeneous frequency responses. In this regard, agents with short investment horizons,
such as day traders or hedge funds, are more concerned about the short-run performance
of markets and make decisions largely based on ephemeral phenomena like sporadic
events and psychological factors. Hence, their reaction to shocks occurs principally in the
short-run. Meanwhile, other agents, such as big institutional investors, are more interested

4
in the long-run market performance, so that their response to economic and financial
shocks is mostly materialized in the long-run. Consequently, it seems reasonable to
assume the existence of linkages with various degrees of persistence and, hence, different
frequency sources of connectedness among oil and financial markets.
To our best knowledge, this study is the first to investigate return and volatility
connectedness across clean energy stock prices, crude oil prices and several major
financial variables at both time and frequency domains simultaneously using the time-
frequency connectedness method of Barunik and Krehlik (2018). Actually, the recent
study of Ahmad (2017), which is the most closely related to the present research, uses the
Diebold-Yilmaz framework to explore the existence of time-varying spillovers between
returns and volatilities of crude oil prices and stock prices of U.S. technology and
renewable energy firms. Beyond this methodological issue, our research differs from that
of Ahmad (2017) in two relevant respects. Firstly, we examine connectedness by
considering a larger set of variables. In addition to the variables employed by Ahmad
(2017), our analysis includes the U.S. conventional energy sector index, U.S. 10-year
Treasury bond yields, the U.S. default spread, and the implied volatility of the U.S. stock
and Treasury bond markets, as measured by the VIX and TYVIX indices, respectively.3
Secondly, we extend the time period covered by Ahmad (2017) by using a dataset from
January 2003 to September 2017. This sample fully accounts the recent collapse of crude
oil prices by more than 70%, from highs of $106 in July 2014 to $27 in February 2016,
as well as the post-collapse months, when a slight recovery has been observed.
The key findings of this study can be summarized as follows. First, return and volatility
connectedness among the variables under examination comes primarily from the higher
frequency band (up to five days). This indicates that crude oil and financial markets have
become highly efficient and shocks get transmitted very quickly across markets, causing
responses shorter than one week. Second, an increase in the degree of connectedness is
observed since the onset of the U.S. subprime mortgage crisis in summer of 2007,
consistent with the notion that interconnectedness among crude oil and financial markets
rises substantially during periods of financial turmoil. In the face of widespread fear and
uncertainty in markets, any new information is scrutinized and processed more carefully,
3 To our best knowledge, Auran and Gullaksen (2017) is the only study until date that has analyzed the
impact of VIX returns on equity prices of renewable energy firms. However, using a time-varying
multifactor model, these authors conclude that VIX is the least influential factor, exerting little or no
influence on renewable energy stock prices.

5
thus generating heightened connectedness. This result is also in agreement with a general
re-pricing of risk triggered by the 2007-2008 global financial crisis. Third, crude oil prices
are not identified as a major driver of the stock market performance of renewable energy
stocks in the short-term or the long-term. This finding suggests a decoupling of the clean
energy industry from the traditional energy market, in line with the idea that the
alternative energy is becoming its own sector independent of the fossil fuel energy sector.
As a matter of fact, crude oil prices emerge as a net receiver of financial shocks during
the entire sample period, which is consistent with the financialization process of the oil
commodity market that has taken place since the early 2000s. Fourth, a significant
pairwise volatility connectedness is found, mainly at the higher frequency band, between
clean energy and technology stock prices over most of the sample. This supports the view
that investors perceive green energy companies as similar to high technology companies.
Finally, traditional energy stock prices seem to contain some relevant information
concerning the evolution of crude oil prices in the very short-term.
Connectedness results in this paper may be useful for several economic agents with
different investment horizons in order to implement better portfolio diversification and
hedging strategies and optimal policy measures. For short-term investors and portfolio
managers, it is really complicated to construct well diversified portfolios consisting of
U.S. alternative and conventional energy stocks, technology stocks, 10-year Treasury
bonds, VIX and TYVIX futures, and crude oil-related assets, mainly in times of financial
turmoil.4 In particular, the combination of large proportions of high tech and renewable
energy stocks in a portfolio is especially unadvisable for agents with short horizons.
Meanwhile, the inclusion of crude oil-related assets in portfolios composed primarily of
clean energy and technology stocks can provide interesting long-term diversification
benefits to investors and portfolio managers with longer time horizons. In addition, for
policy makers, the lack of a meaningful connectedness between crude oil prices and
alternative energy stock prices implies that the alternative energy sector does not need
specific policies of protection against the short- and long-term impact of crude oil price
fluctuations.
The rest of the paper proceeds as follows. Section 2 presents a brief review of the relevant
literature on the interactions between crude oil prices and stock prices of clean energy
4 Investors cannot invest directly in volatility indices but can invest in volatility derivatives. VIX and
TYVIX futures are traded on the CBOE Futures Exchange (CFE) since 2004 and 2014, respectively.

6
companies. Section 3 formally introduces the econometric framework used in this study,
while Section 4 describes the dataset employed. Section 5 reports and discusses the most
significant empirical results. Finally, Section 6 provides some concluding remarks.
2. Literature review
The vertiginous growth of the renewable energy sector in recent years has given rise to
an emerging body of research focused on the link between the stock market performance
of alternative energy companies and crude oil prices. Henriques and Sadorsky (2008)
were the first to investigate the dynamic relationship between clean energy stock prices
and oil prices using a vector autoregression (VAR) model. These authors find significant
Granger causal effects from technology stock prices to renewable energy stock prices.
However, the impact of oil price shocks on stock prices of sustainable energy firms is
limited. Later studies by Kumar et al. (2012) and Managi and Okimoto (2013) extended
the paper of Henriques and Sadorsky (2008) by considering the price of European Union
carbon allowances before and after the oil price peak in July 2008 and allowing for
structural breaks, respectively, in the VAR framework. In contrast to the findings of
Henriques and Sadorsky (2008), Kumar et al. (2012) and Managi and Okimoto (2013)
document a significant positive response of clean energy stock prices to oil price shocks,
suggesting the presence of a substitution effect between traditional fossil fuels and
renewable energy. Furthermore, the three above mentioned studies consistently show that
shocks to technology stock prices have a larger positive effect on clean energy stock
prices than do oil prices. This finding is in accordance with the idea that alternative energy
stocks are viewed by investors as very similar to high technology stocks.
Subsequent works have further elaborated on the existing connection between oil and
clean energy stock prices employing different econometric approaches. For example,
using a multivariate GARCH (Generalized Autoregressive Conditional
Heteroskedasticity) framework, Sadorsky (2012) finds that stock prices of renewable
energy firms correlate more highly with technology stock prices than with oil prices in
terms of volatility. More recently, Inchauspe et al. (2015) and Reboredo (2015) document
that the relationship between oil prices and the returns of alternative energy companies is
time varying. To that end, Inchauspe et al. (2015) estimate a multifactor asset pricing
model with time-varying coefficients, while Reboredo (2015) uses dynamic copulas.
Reboredo (2015) further shows that oil price movements contribute around 30% to

7
renewable energy systemic risk based on as the conditional Value-at-risk (CoVaR)
measure.
Allowing for the presence of endogenous structural breaks, Bondia et al. (2016) reveal
the existence of a long-run cointegrating relationship among stock prices of alternative
energy and tech companies, oil prices and interest rates. They also report evidence of
short-run Granger causality running from the three later variables to clean energy stock
prices, while there is no evidence of causality running towards renewable energy stock
prices in the long run. In turn, Reboredo et al. (2017) study co-movement and causality
between oil and green energy stock prices using continuous and discrete wavelet methods.
Their empirical results indicate that dynamic interactions between crude oil prices and
new energy returns are weak in the short run, but gradually increase over the long run.
They also find linear causality at lower frequencies, while nonlinear causality is detected
at different time horizons running mostly from clean energy stock returns to oil prices. In
another interesting contribution, Dutta (2017) shows that oil price uncertainty, as
measured by the oil price implied volatility index (OVX), is a highly significant variable
for modeling and forecasting the volatility of alternative energy stock returns, particularly
during the global financial crisis period. Lastly, using the spillover index approach
proposed by Diebold and Yilmaz (2012), Ahmad (2017) shows that technology stocks
play a crucial role as a transmitter of return and volatility spillovers to renewable energy
stocks and oil prices. Meanwhile, crude oil price exhibits limited interdependence with
clean energy and technology stocks and is a net receiver of spillovers.
Overall, the empirical findings of this rapidly growing body of literature suggest that there
is no consensus among economists about the relationship between crude oil prices and
stock prices of clean energy companies. However, despite using a wide variety of
methodological approaches, most of these studies conclude that the stock market
performance of renewable energy firms is strongly correlated with that of high technology
firms, while oil prices play generally a minor role in comparison.
3. Empirical methodology
This paper employs the connectedness methodology introduced by Barunik and Krehlik
(2018) to assess the dynamic interactions in time and frequency among stock prices of
renewable energy companies, crude oil prices and various influential financial variables.
This technique can be viewed as an extension to the time-frequency space of the better
known spillover index approach put forward by Diebold and Yilmaz (2012).

8
The Diebold-Yilmaz method is based on the notion of forecast error variance
decomposition within the generalized VAR framework proposed by Koop et al. (1996)
and Pesaran and Shin (1998) to estimate the magnitude and direction of connectedness in
the time domain. However, Barunik and Krehlik (2018) expand the Diebold-Yilmaz
approach by including the spectral representation of variance decompositions (e.g., Dew-
Becker and Giglio, 2016; Stiassny, 1996), which allows estimating unconditional
connectedness relations in the frequency domain. Therefore, the distinctive feature of the
Barunik-Krehlik framework is its ability to measure the dynamics of connectedness
among a set of variables over time and across different frequencies simultaneously. In
any case, before introducing the time-frequency connectedness measures, we briefly
discuss the major features of the Diebold-Yilmaz method.
A relatively simple and effective way to quantify connectedness in the time domain is to
consider a VAR process and to compute its forecast error variance decomposition. Hence,
let us have a VAR model with n variables and p lags, which can be written as:
=()+
(1)
where  denotes a × 1 vector of endogenous variables, ()=
 is a × p-
th order lag polynomial matrix of coefficients, L is the lag operator and  represents a
white noise error vector with zero mean and covariance matrix Σ.
Assuming covariance stationarity, the moving average representation of the VAR process
is given by:
=()=


+
(2)
where () is a × matrix of infinite lag polynomials that can be calculated
recursively.
Under the generalized VAR identification scheme proposed by Pesaran and Shin (1998),
the generalized forecast error variance decomposition of a variable into components
attributable to shocks to the different variables in the system for a forecast horizon H can
be computed as:
()=
 ()


(
)

 ,
(3)

9
where  stands for a × matrix of moving average coefficients corresponding to the
lag h,  is the jth diagonal element of the Σ matrix and H is the selected forecast horizon.
The () denotes the contribution of the jth variable to the variance of forecast error of
the variable ith at the horizon H.
In the generalized VAR framework, the sum of own- and cross-variable variance
contribution shares is not necessarily equal to one, i.e. ()

 1. Hence, each
entry of the variance decomposition matrix can be normalized by the row sum as s:

()=()
 ()


(4)
As can be seen, 
() provides a measure of pairwise connectedness from j to i at
horizon H in the time domain. Moreover, using the above defined normalized variance
contribution, a number of measures which reflect the degree of connectedness among the
variables of the system can be conveniently introduced.
As argued by Diebold and Yilmaz (2012), generalized forecast error variance
decompositions are crucial for assessing connectedness in the time domain. Analogously,
to describe connectedness in the frequency domain, it is necessary to consider the spectral
representation of the variance decomposition based on frequency responses to shocks
instead of impulse response to shocks. Following the approaches of Dew-Becker and
Giglio (2016) and Stiassny (1996), spectral decomposition methods are used to explore
connectedness relationships in the frequency domain. In this setting, the frequency
response function plays a vital role. This function, which can be obtained as the Fourier
transform of the coefficients , with =1, can be defined as:
= 


(5)
where  denotes the frequency.
In turn, the power spectrum, (), which shows how the variance of  is distributed
over the frequency components , is given by:
()=()


=
(6)
As noted by Krehlik and Barunik (2017), using the spectral representation for the
covariance, the frequency response functions can be employed to derive the generalized

10
variance decompositions in the frequency domain. Specifically, the generalized forecast
error variance decomposition at a particular frequency  can be computed as:
()=
 



()()


(7)
where () represents the portion of the spectrum of the variable ith at a given
frequency  that can be attributed to shocks in the variable jth. As can be seen, the
forecast horizon H does not play a significant role in this context.
As in the case of the time domain analysis, Eq. (7) can be normalized as:

()=()
()


(8)
Importantly, 
() measures pairwise connectedness from j to i at a given frequency 
and, therefore, it can be interpreted as a within-frequency causality indicator. In contrast,
the above mentioned 
() reflects pairwise connectedness from j to i at a particular
horizon H, so that it can be viewed as an indicator of the strength of causality exclusively
in the time domain. In this regard, when Diebold and Yilmaz (2012) quantify
connectedness using 
(), they focus on information aggregated through frequencies,
while the possible heterogeneous frequency responses to shocks are completely ignored.
In economic and financial applications, it can be interesting to assess short-, medium- or
long-term connectedness rather than connectedness at a single frequency. Hence, it seems
more appropriate to work with frequency bands. In this setting, the accumulative
connectedness at an arbitrary frequency band =(,) can be obtained as:

() = 
()


(9)
From here, it is possible to define a variety of connectedness measures in the frequency
domain, which are inspired by the indicators introduced by Diebold and Yilmaz (2012)
for the time domain. For example, the overall connectedness within the frequency band d
can be calculated as:
=
()

,

()
 = 1 
()



()

(10)

11
Note that  close to unity implies strong connections within the frequency band d.
However, the weight of this spectral band within the aggregate connectedness might be
very low. Hence, these measures are called within measures as they describe only the
connectedness within a particular frequency band.
As pointed out by Krehlik and Barunik (2017), the Barunik-Krehlik framework also
allows the identification of the direction of spillovers. Specifically, the portion of variance
of variable i contributed by all the other variables   , which is called within from
connectedness, at the frequency band d can be computed as:

= 
()

,
(11)
Analogously, the contribution of variable i to all the other variables j (  ) is called
within to connectedness on the spectral band d and is given by:

= 
()

,
(12)
In addition, the so-called within net connectedness, which quantifies the difference
between the variance transmitted and received by a given variable, is defined as:
,
=


(13)
If ,
 is positive, it means that the variable i transmits more information than it receives
from the remaining elements of the system.
Apart from the system-wide behavior, it can be very useful to investigate pairwise
relationships. Thus, the net pairwise connectedness between two variables i and j can be
calculated as:

=
()
()
(14)
According to Barunik and Krehlik (2018), to get an indicator of the contribution of a
given frequency band to the aggregate connectedness, the within measures have to be
weighted. Thus, the contribution of the frequency band d to the overall system
connectedness can be obtained as:

=()
(15)
where the spectral weight ()=
()

 



=
()

 
 reflects the
contribution of the frequency band d to the whole VAR system, while  is the total

12
connectedness measure corresponding to the spectral band d computed according to Eq.
(10). Lastly, it should be mentioned that the sum of all frequency connectedness measures
over disjointed intervals is equal to the original total connectedness measure proposed by
Diebold and Yilmaz (2012), i.e., =

.
4. Data
The dataset for this study consists of daily closing prices of U.S. renewable energy stocks,
high technology stocks, conventional energy stocks, crude oil futures contracts, U.S. 10-
year Treasury bond yields, the default spread and the volatility of the U.S. stock and
Treasury bond markets. All the data are gathered from Thomson Reuters DataStream,
excepting the default spread, which is obtained from the Federal Reserve Bank of St.
Louis. The sample period covers January 2, 2003 until September 29, 2017, containing a
total of 3724 daily observations. The variables are described below.
4.1. Description of variables
The Wilder Hill Clean Energy Index (ECO) is used to measure the stock market
performance of U.S. alternative energy firms. ECO was the first index created to track
the stock prices of renewable energy companies and has become a benchmark index in
this field. It is an equal-dollar-weighted index consisting of a set of corporations engaged
in activities related to the use of cleaner forms of energy, such as solar power, wind power,
hydrogen and fuel cells, biofuels, pollution prevention and related areas. This index is
quarterly rebalanced and it is composed of 40 companies in the first quarter of 2018. ECO
has been employed in the vast majority of studies on the link between new energy stock
prices and oil prices (Ahmad, 2017; Bondia et al., 2016; Dutta, 2017; Henriques and
Sadorsky, 2008).
The NYSE Arca Tech 100 Index, which was formerly known as the Pacific stock
exchange (PSE) index and still maintains the ticker symbol PSE, is one of the oldest U.S.
technology-based stock market indices. PSE is a price-weighted index composed of
common stocks and ADRs of technology-using companies operating across a broad
spectrum of industries (e.g. computer hardware, software, semiconductors, aerospace and
defense, and biotechnology) and listed in the U.S. stock exchange. The consideration of
high technology equities in this study stems from the view that clean energy companies
may have more in common with tech companies than even they do with the crude oil
market. As pointed out by Sadorsky (2012), the success or failure of alternative energy

13
firms often depends on the level of technical development achieved by technology firms.
Moreover, companies in the renewable energy and technology sectors are competing on
many occasions for the same resources, access to infrastructures and institutional support.
In fact, the behavior of the U.S. equity market during the late 1990s and early 2000s
supports the high similarity of renewable energy and technological stocks. It should be
noted that the PSE has been also used in many studies in this field (Inchauspe et al., 2015;
Kumar et al., 2012; Managi and Okimoto, 2013; Sadorsky, 2012). In turn, the S&P 500
Oil, Gas & Consumable Fuels sector index (CESI) is utilized to account for the equity
market performance of U.S. conventional energy firms. CESI is made up of companies
active in integrated Oil and Gas, Oil and Gas exploration and production, Oil and Gas
refining and marketing, Oil and Gas storage and transportation, and coal and consumable
fuels. Therefore, CESI can be considered as a major benchmark for the traditional fossil
fuel energy industry.
In this study, we measure oil prices using the closing prices of the nearest contract to
maturity on the West Texas Intermediate (WTI) crude oil futures contract for several
reasons. Firstly, as argued out by Sadorsky (2001), spot oil prices are more heavily
affected by short-run noise due to temporary shortages or surpluses than future oil prices.
Secondly, oil futures contracts are the most widely traded physical commodity in the
world and, hence, they constitute a benchmark for the oil market. Thirdly, WTI represents
the major benchmark for U.S. crude oil. Moreover, crude oil futures contracts have been
employed in this literature by numerous authors, including Ahmad (2017), Henriques and
Sadorsky (2008), Kumar et al. (2012) and Managi and Okimoto (2013). Interest rates are
typically identified as a significant explanatory factor of the behavior of the overall stock
market (Jareño et al., 2016; Moya-Martínez et al., 2015; Reilly et al., 2007). In this way,
interest rates may play a role in explaining movements in stock prices of new energy
companies (see e.g., Bondia et al., 2016; Henriques and Sadorsky, 2008; Kumar et al.,
2007; Managi and Okimoto, 2013). In this context, the yield on U.S. 10-year Treasuries
is also considered in this research. The use of the 10-year government bond yield has
become very frequent in the literature on the interest rate-stock market nexus. As
discussed by Ferrer et al. (2016), 10-year interest rates incorporate market expectations
about future prospects for the economy and largely determine the cost of borrowing. Thus,
long-term interest rates are likely to have a critical influence on investment decisions and
profitability of firms and, hence, on their stock market performance.

14
In addition, the default spread, defined as the difference between Moody’s seasoned
yields on Baa and Aaa corporate bonds, is also considered. This spread is a proxy for
aggregate credit or default risk in the economy. Changes in the default spread can contain
important information about the evolution of the real economy. An expected
improvement in economic conditions leads to lower credit risk and a narrowing of the
default spread. Instead, deteriorating economic prospects at the aggregate level lead to
higher credit risk and a widening of the default spread. Studies such as Bhardwaj et al.
(2015), Gilchrist and Mojon (2014) and Gilchrist and Zakrajšek (2012) have shown the
predictive power of default spreads for economic activity. Accordingly, we include the
default spread in this study to account for the effect of business cycle movements on the
connectedness among U.S. clean energy stock prices, crude oil prices and the selected
financial variables.
Finally, our dataset includes two volatility indices calculated by the Chicago Board of
Options Exchange (CBOE), namely VIX and TYVIX. VIX measures the market’s
expectation of volatility of the S&P 500 stock index over the next 30 calendar days based
on the market prices of options on the S&P 500 index spanning a wide range of strikes.5
VIX, often referred to as the fear index, is broadly recognized as an indicator of investors’
risk aversion. A surge in VIX is a sign of greater uncertainty and fear in the stock market
and increased risk aversion of investors. In turn, TYVIX is intended to capture investors’
expected volatility of the U.S. 10-year Treasury note futures traded on the Chicago Board
of Trade (CBOT) over a 30-day horizon by applying the VIX methodology. Thus, TYVIX
represents the expected future volatility of the Treasury bond market as VIX reflects the
expected volatility of the stock market. The TYVIX index provides a useful indicator of
changes in implied volatility in the U.S. Treasury market. The more uncertainty there is
around changes in interest rates, the higher TYVIX is. It is worth noting that no prior
study has examined the effect of TYVIX on stock prices of clean energy companies.
4.2. Preliminary results
Figure 1 plots the evolution over time of each of the eight selected variables. As can be
seen, all the variables exhibit large variability over the period under study. A clear upward
trend is observed for ECO, PSE, CESI and crude oil futures contracts from the beginning
of 2003 until the intensification of the financial crisis in autumn of 2008. Interestingly,
5 The exact formula and the calculation process for VIX can be found at
https://www.cboe.com/micro/vix/vixwhite.pdf.

15
two distinct stages are identified in the linkage between alternative energy stock prices
and oil prices during the full sample. ECO and crude oil futures follow a very similar
increasing trend during the early part of the sample period and suffer a big drop in mid-
2008. However, they behave differently in the post-financial crisis period as the
performance of the ECO index becomes more independent of the ups and downs of crude
oil prices. In addition, it is shown that PSE has recovered much faster from the financial
turbulence in late 2008 and the subsequent economic downturn than ECO. After a
relatively vigorous performance between 2003 and 2007, ECO seems to be
underperforming after the worldwide financial crisis in comparison to high technology
stocks. Similarly, CESI has also recovered better from the financial crisis than ECO,
although the performance of CESI has been seriously harmed by the substantial decline
in crude oil prices from mid-2014 to early-2016. In turn, U.S. 10-year Treasury bond
yields display a slight rising trend from early 2003 to mid-2007, coinciding with the first
symptoms of the U.S. subprime mortgage crisis. From here, yields on 10-year Treasuries
follow a marked declining trend, which is exacerbated during the financial turmoil in late
2008. Lastly, DEF, VIX and TYVIX are also profoundly impacted by the global financial
crisis, reaching historic records in the weeks after the bankruptcy of Lehman Brothers in
September 2008.
In order to ensure the stationary of the data, stock and crude oil returns are calculated as
daily logarithmic returns. Meanwhile, following the usual practice, movements in U.S.
10-year Treasury bond yields, changes in the U.S. default spread and changes in VIX and
TYVIX are computed as the first differences between two successive observations. Table
1 provides summary statistics for all the series over the entire sample. The average daily
returns for PSE, CESI and crude oil futures are positive, although the highest average
return is found for PSE, confirming the overall upward trend of high technology stock
prices. In contrast, the average daily return for ECO is negative, mainly due to the poor
stock market performance of clean energy firms from the onset of the international
financial crisis. Additionally, the average daily changes in 10-year Treasury bond yields,
the default spread and the VIX and TYVIX indices are negative and very close to zero.
In terms of standard deviations, the oil market is more volatile than the stock market
(renewable and traditional energy, and technology sectors) and the government debt
market as a result of the enormous price swings of crude oil over the last fifteen years.
Inchauspe et al. (2015) also document higher volatility for the crude oil market than for

16
clean energy and technology stock indices and the MSCI world stock index for the time
period September 2001 to February 2014, which does not account for the recent collapse
of crude oil prices.
As expected, TYVIX has a smaller standard deviation than VIX, reflecting the lower risk
of the long-term Treasury debt market in comparison to the equity market. There is no
clear pattern of skewness in the data. The skewness measure is negative for ECO, PSE,
CESI and changes in yields on 10-year Treasury bonds, while it is positive for the rest of
series. The kurtosis coefficient is always greater than 3, indicating that all the series have
heavy tails and peakedness relative to the Gaussian distribution. This departure from
normality is supported by the Jarque-Bera test statistics, which reject the null hypothesis
of a normal distribution for all the series at the 1% significance level. Lastly, the results
of the Augmented Dickey-Fuller (ADF) unit root test and the Kwiatkowski-Phillips-
Schmidth-Shin (KPSS) stationarity test reveal that all the series are stationary processes,
i.e. I(0), at the 1% level.
INSERT TABLE 1 ABOUT HERE
Table 2 reports unconditional correlation coefficients between all pairs of variables over
the whole sample. In line with Ahmad (2017), Inchauspe et al. (2015) and Sadorsky
(2012), the highest magnitude of the unconditional correlation is observed between ECO
and PSE, with a positive value of 0.79, implying that there is a strong comovement
between equity prices of renewable energy and high technology companies. In fact, this
correlation is more than double than that found between ECO and crude oil prices (0.30).
Not surprisingly, a high positive correlation (0.51) is also found between CESI and crude
oil prices, reflecting that the profitability and stock valuation of conventional energy-
related firms are positively dependent on the price of crude oil (e.g., Boyer and Filion,
2007; Junttila et al., 2018; Moya-Martínez et al., 2014). Moreover, the unconditional
pairwise correlations between ECO and VIX, PSE and VIX, and CESI and VIX take large
negative values (-0.67, -0.76 and -0.69, respectively). This strong negative association
means that fear and uncertainty in the stock market have a detrimental effect on clean
energy, traditional energy and technology stock prices. Overall, the high unconditional
pairwise correlations here presented suggest the possible existence of important
interactions among the variables under scrutiny, thus stressing the advisability of carrying
out a return and volatility connectedness analysis.

17
Finally, the conditional volatility of each series to be employed in the analysis of volatility
connectedness is estimated using a univariate conditional volatility model. A number of
GARCH processes, including the standard symmetric GARCH(1,1) and several popular
asymmetric GARCH(1,1) models, such as the EGARCH, GJR-GARCH and APARCH
specifications, under various alternative conditional distributions (i.e. Gaussian, Student-
t and Generalized error distribution along with their respective skewed versions), are
considered to select the best GARCH model. In particular, the best-fit GARCH(1,1)
process for each series is the one which minimizes the negative log-likelihood function.6
INSERT TABLE 2 ABOUT HERE
5. Empirical results
This section presents the estimation results of the return and volatility connectedness
measures based on the time-frequency approach put forward by Barunik and Krehlik
(2018). Given that the central aim of this paper is to investigate the time and frequency
dynamics of interactions among U.S. clean energy stock prices, WTI oil prices and
several influential financial indicators, we focus on the dynamic version of the
connectedness method of Barunik and Krehlik (2018). Following Krehlik and Barunik
(2017), connectedness measures are calculated in this paper for two different frequency
bands. The first spectral band corresponds to movements up to five days (one week),
while the second one refers to movements from six to two hundred days. These frequency
bands enable one to capture the short- and long-term dynamic connectedness,
respectively, among crude oil prices, renewable energy stock prices and the rest of
financial indicators. In particular, the dynamic connectedness measures are estimated
using a rolling window size of 200 days and a forecast horizon of H=100 days.7 For each
rolling window, the optimal lag length of our eight variable VAR system is chosen
according to the Schwarz information criterion. The results of the pure time-domain
approach of Diebold and Yilmaz (2012) are also reported for comparison purposes.
5.1. Total return and volatility connectedness
6 For the sake of brevity, the estimation results of the GARCH models are omitted, but they are available
from the authors upon request.
7 The choice of a forecast horizon of 100 days is based on the original paper by Barunik and Krehlik (2018),
although the connectedness approach developed by these authors is not affected by the selected forecast
horizon.

18
Figure 2 illustrates the time dynamics of the total return and volatility spillover indices
computed based on the Diebold-Yilmaz framework. Total spillovers are relatively high
during most of the sample period, indicating a significant degree of connectedness among
stock prices of alternative and traditional energy companies and tech companies, crude
oil prices and the remaining financial variables over the full sample. The system-wide
return and volatility spillover indices exhibit a rather similar pattern, although the
magnitude of return spillovers tends to be slightly larger than that of volatility spillovers
during almost all the sample period.
Beginning with fairly stable values around 40% and 20% for the total return and volatility
spillover indices, respectively, during mid-2000s, both total spillover measures went up
substantially from the first quarter of 2007 following the first symptoms of the crisis in
the U.S. subprime mortgage market. As expected, the Lehman Brothers bankruptcy on
September 15, 2008 and the associated financial meltdown led to a further increase in
connectedness during the last quarter of 2008. After a short temporary decrease, total
connectedness went back up considerably in late 2010 during the most significant period
of contagion of the Eurozone sovereign debt crisis. In particular, the historic peak of total
return connectedness (64.84%) is reached in November 2010, while the high record of
total volatility connectedness (58.89%) is reached in October 2008. These findings clearly
show the enormous impact of the 2007-2008 financial crisis and the subsequent European
sovereign debt crisis on both return and volatility spillovers. This supports the widespread
view that interactions among commodities and financial markets rise drastically during
periods of heightened financial turmoil (Bhardwaj et al., 2015; Creti et al., 2013; Krehlik
and Barunik, 2017; Li et al., 2016). When uncertainty reigns, any positive or negative
information is reviewed and processed more thoroughly, thus generating increased
interconnectedness.
Furthermore, it is worth highlighting that, from these dates until the end of the sample,
the total spillover indices have stayed at higher levels than during the early part of the
sample period, which is in line with the evidence of Junttila et al. (2018). The
strengthening of spillover effects mainly from the aggravation of the financial crisis in
autumn of 2008 is consistent with the idea that increased economic and financial
uncertainty and risk aversion in the backdrop of the global financial crisis led to a general
re-pricing of risk in the world economy. In this way, a scenario of greater
interconnectedness has emerged in the wake of the recent global financial crisis in which

19
economic agents have become more aware of the vulnerability of the economy to internal
and exogenous shocks. As a result, closer attention is paid to the evolution of the major
macroeconomic and financial indicators, which has ended up materializing in higher
levels of spillovers. This finding is also in accordance with that of Ahmad (2017), who
documents a sharp rise in connectedness between stock prices of new energy and
technology firms, and crude oil prices following the 2007-2008 international financial
crisis. In addition, it should be noted that total spillover indices, especially volatility
spillovers, experienced a drop during the year 2014. This decline in total connectedness
could be related to the crude oil price collapse in July 2014, which was caused by a
combination of factors. On the one hand, business cycle effects, especially the economic
slowdown in China, India and other emerging countries. On the other hand, supply side
factors such as the dramatic growth in North American oil production mostly driven by
Canadian oil sands and U.S. shale oil, and the Saudi Arabia’s refusal to cut crude oil
production.
INSERT FIGURE 2 ABOUT HERE
Figure 3 decomposes the dynamics of total return and volatility connectedness into higher
(up to 5 days) and lower (from 6 to 200 days) frequency bands based on the time-
frequency approach of Barunik and Krehlik (2018). In other words, this figure breaks
down the overall time-varying return and volatility spillover indices provided by Figure
2 into the higher and lower frequencies. Regarding the rolling windows, the window size
is kept in 200 days and the forecast horizon in 100 days. The data points plotted in each
graph reflect total return or volatility connectedness associated with the higher or lower
frequency band for each window, corresponding their position to the end of each window.
For instance, the first windows covers the period from January 2, 2003 to October 15,
2003 (with a total of 200 daily observations). Regarding the results, it is evident that the
largest portion of interconnectedness in terms of both return and volatility occurs at the
higher frequency band during the whole sample period. As a matter of fact, the time
dynamics of connectedness found for the frequency band of up to 5 days (one week) is
very similar to that resulting from the standard Diebold-Yilmaz methodology. Thus, a
steep increase in total return and volatility connectedness is observed since the end of the
first quarter of 2007, in parallel with the first signs of the U.S. subprime crisis. The
historic maximums of return and volatility connectedness are reached in late 2008 during
the most acute phase of the worldwide financial crisis and also during the intensification

20
of the European debt crisis in late 2010, confirming that connectedness among crude oil
and financial markets increases considerably in times of financial turmoil.
The prevalence of the higher frequency band suggests that total connectedness among
alternative energy stock prices, crude oil prices and the selected financial indicators is
mostly driven by the transmission of shocks in the very short-term. This implies that
markets process information rapidly, so that shocks to any asset or market are transmitted
to other assets and/or markets basically within one week.8 It is shown that at the higher
frequency band the magnitude of return connectedness is almost always greater than that
of volatility connectedness, indicating that in the short-term return spillovers have a
stronger impact on the system than volatility spillovers. With respect to the lower
frequency band, we also generally observe a higher degree of connectedness in terms of
return than in terms of volatility. However, long-term volatility spillovers are larger than
long-term return spillovers during the most severe stages of the global financial crisis and
the European debt crisis. This greater relative importance of volatility spillovers in the
long-term suggests that the impact of volatility shocks is magnified during episodes of
high financial stress such as those associated with the recent crises.
INSERT FIGURE 3 ABOUT HERE
5.2. Net directional return and volatility connectedness
The next step in the analysis is to focus on directional information in order to identify the
main net transmitters and receivers of spillovers. The time-varying net directional return
and volatility spillovers across all the variables of the system calculated using the pure
time-domain approach of Diebold and Yilmaz (2012) are plotted in Figures 4 and 5,
respectively. By concentrating on net spillovers, it is possible to discern whether a
particular variable is a net transmitter or a net receiver of spillover effects to/from all the
remaining variables. A positive value of the net directional spillover index means that the
variable in question is a net transmitter of spillovers to all the other variables, while a
negative value indicates that such variable is a net receiver of spillovers.
A quite similar pattern regarding the major net transmitters and receivers of shocks is
observed for directional return and volatility spillovers, although volatility spillovers
8 It is worth highlighting that additional intermediate frequency bands, e.g. from 6 days to 3 months, have
also been considered. However, the results revealed that most of return and volatility connectedness
concentrate at the higher frequency band (up to 5 days). Accordingly, we report only the empirical results
for the original higher (up to 5 days) and lower frequencies (from 6 to 200 days).

21
show a greater level of variability than return spillovers. Just as the previous total
connectedness measures, net directional spillovers exhibit a considerable time variation,
with the highest values being generally reached at the height of the worldwide financial
crisis and the subsequent Eurozone sovereign debt crisis. Importantly, stock prices of
conventional and clean energy, and technology companies together with VIX emerge as
net transmitters of return and volatility spillovers to all the other variables over most of
the sample period. In contrast, WTI crude oil futures, yields on U.S. 10-year Treasuries,
the U.S. default spread and the volatility of 10-Treasury bond yields, as measured by
TYVIX, are found to be net receivers of return and volatility spillovers from all the
remaining variables for the longest part of the period of study. This result can be
interpreted as a first sign that crude oil prices are not a key determinant of the equity
market performance of alternative energy companies, which suggests a decoupling of the
green energy sector from the oil market over the last years. Conversely, crude oil prices
are significantly influenced by the behavior of the remaining variables for the most part
of the sample period, which can be associated with the financialization of the oil market
that has occurred since the early 2000s.
INSERT FIGURE 4 ABOUT HERE
INSERT FIGURE 5 ABOUT HERE
Figures 6 and 7 depict the time-frequency dynamics of the net directional return and
volatility connectedness, respectively, estimated using the methodology of Barunik and
Krehlik (2018). Again, the results for return and volatility connectedness are rather
similar to each other, although a higher variability is observed for volatility
connectedness. The values of the net directional connectedness measures reveal that the
linkages between each individual variable and the rest of variables in the system tend to
be more pronounced at the higher frequency band (up to 5 days) than at the lower
frequency band (from 6 to 200 days) during the entire sample period. This finding is fully
in line with the evidence about total connectedness given by Figure 3, confirming that
interconnectedness between financial and crude oil markets mostly occurs in the very
short-term, i.e. up to one week. Furthermore, the record high level of net directional
connectedness is reached for virtually all the variables at the heart of the global financial
crisis or the European debt crisis, demonstrating once again that the degree of
interdependence among commodity and financial markets greatly enhances in times of
financial turmoil.

22
As can be seen, stock prices of high technology, conventional and renewable energy
companies and VIX appear as the principal net transmitters of return and volatility
connectedness to all the other variables both at the higher and lower frequency bands over
most of the period under review. However, it is worth mentioning that clean energy stock
prices were net receivers of return and volatility spillovers at the lower frequency band
during the period preceding the intensification of the financial crisis in late 2008. One
possible explanation for the changing role of green energy stocks could be related to the
deep transformation of the global energy market from mid-2008. More concretely, until
the bursting of the oil price bubble in the second half of 2008, the long-term equity market
performance of alternative energy firms was basically driven by the development of crude
oil prices. Since then, however, alternative energy has become a less dependent sector of
the traditional energy market, and hence renewable energy stock prices have followed a
quite different path than crude oil prices in recent years. The time-varying nature of the
relationship between oil prices and clean energy stocks has been also evidenced by many
studies, such as Managi and Okimoto (2013), Sadorsky (2012), Inchauspe et al. (2015)
and Reboredo et al. (2017).
On the contrary, WTI oil futures contracts, U.S. 10-year Treasury bond yields, the U.S.
default spread and the volatility of the U.S. government bond market are net receivers of
return and volatility connectedness from all the other variables during most of the sample
period both at the short- and the long-term. In fact, this pattern only seems to change
during the most acute stages of the worldwide financial crisis and the euro area debt crisis,
when yields on 10-year Treasuries, the default spread and TYVIX play temporarily the
role of net transmitters of volatility spillovers, primarily in the short-term.
INSERT FIGURE 6 ABOUT HERE
INSERT FIGURE 7 ABOUT HERE
5.3. Pairwise directional return and volatility connectedness
Next, we shift our attention to the pairwise directional spillovers to shed some light on
the key transmitters and receivers of shocks in a bivariate setting. Figures 8 and 9 plot the
network graphs of the average net pairwise directional return and volatility spillovers
estimated using the time-domain Diebold-Yilmaz approach, respectively, providing an
intuitive and direct visualization of the level of pairwise connectedness. The whole
sample period is divided into three sub-periods to analyze the evolution over time of

23
return and volatility connectedness. In particular, the network diagrams refer to the three
following sub-periods: pre-financial crisis (from January 2003 to July 2007), financial
crisis (from August 2007 to July 2012) and post-financial crisis (from August 2012 to
September 2017).9 To facilitate the identification of the main patterns of
interconnectedness, the plots in these figures only display the most significant net
pairwise linkages between the eight variables of our system.
Several important results arise from this analysis. First, stock prices of high technology
and renewable energy companies and VIX emerge as net transmitters of pairwise
directional return and volatility spillovers to the rest of variables regardless of the sub-
period. In contrast, WTI oil futures, U.S. 10-year Treasury bond yields, the volatility of
the U.S. Treasury market and the U.S. default spread act as net recipients of pairwise
spillovers in return and volatility during the full sample period. This evidence is entirely
consistent with the findings regarding net total directional connectedness presented in
Figures 4 and 5. Second, it is also noteworthy the dramatic change of role experienced by
stock prices of conventional energy companies. More precisely, CESI has moved from
being a net taker of pairwise return and volatility spillovers from VIX, ECO and PSE
during the first sub-period to becoming a net transmitter of pairwise spillovers to yields
on U.S. 10-year Treasuries, the default spread, TYVIX and, mainly, to crude oil prices
since the start of the global financial crisis. The larger influence of CESI since the onset
of the financial tsunami can be framed within the increased concern about risk induced
by the worldwide financial crisis. In this environment of growing interdependence across
markets, a possible explanation for the strong information transmission from CESI to
crude oil prices is that the stock market performance of the fossil fuel energy industry
reflects quite well the expectations for the future economy and the global demand for
crude oil. Hence, stock prices of conventional energy companies are able to anticipate the
future development of crude oil prices.
Third, the number and intensity of significant net pairwise directional return and volatility
spillovers have increased markedly from the beginning of the U.S. subprime mortgage
crisis in summer of 2007. Moreover, many of these linkages have remained during the
post-financial crisis period, in line with the scenario of higher interconnectedness
9 The famous speech by Mario Draghi, president of the European Central Bank (ECB), in July 2012 that
the ECB was willing to do whatever it takes to keep the financial architecture of the Eurozone was a crucial
turning point in the European sovereign debt crisis and has been used in this paper to represent the end of
the financial crisis sub-period.

24
resulting from the recent global financial crisis. Fourth, crude oil futures do not transmit
significant pairwise return and volatility spillovers to equity prices of clean energy
companies regardless of the sub-period. This finding is completely at odds with the
common perception that crude oil prices exert a notable influence on the stock market
performance of alternative energy firms, thus confirming the decoupling of the renewable
energy sector from the crude oil market. In fact, net pairwise return and volatility
spillovers detected between these variables are not very strong and run basically from
green energy companies’ stock prices to crude oil price, thus again reflecting the
financialization of the oil commodity market.
INSERT FIGURE 8 ABOUT HERE
INSERT FIGURE 9 ABOUT HERE
Figures 10 and 11 display the network graphs of the average net pairwise directional
return and volatility connectedness, respectively, based on the time-frequency approach
of Barunik and Krehlik (2018). The average net pairwise connectedness at higher and
lower frequency bands are reported in Panels A and B, respectively, for each of the three
sub-periods above mentioned. As shown in these plots, the net pairwise return and
volatility connectedness measures are much stronger at the higher frequency band (up to
5 days) than at the lower frequency band (from 6 to 200 days) irrespective of the sub-
period. In fact, the pattern of pairwise connections in return and volatility at the higher
frequency band is very similar to that estimated using the pure time-domain Diebold-
Yilmaz framework during the entire sample (see Figures 8 and 9). This finding
corroborates, therefore, the evidence provided by previous figures according to which
most of pairwise directional connectedness takes place within one week, that is, in the
very short-term. A strengthening of net pairwise directional connectedness is also
apparent at both higher and lower frequencies since the outbreak of the U.S. subprime
crisis in mid-2007. Furthermore, this more pronounced interconnectedness tends to be
maintained until the end of the sample, in accordance with the general re-pricing of risk
in the world economy sparked by the financial crisis.
Once again, the results of net pairwise return and volatility connectedness are broadly
similar to each other in terms of the principal transmitters and receivers of shocks. In
particular, equity prices of high technology, traditional and clean energy companies and
VIX are identified as the key transmitters of net pairwise connectedness both in return
and volatility. Interestingly, PSE, CESI and, to a lesser extent, ECO transmit pairwise

25
spillovers to other variables at both higher and lower frequencies during the full sample.
Meanwhile, VIX acts as a net generator of spillovers mainly at the shorter frequency band.
This implies that changes in the level of uncertainty and investors’ risk aversion in the
U.S. stock market, as measured by the VIX fear index, are propagated gradually to other
asset classes, so that markets are slower to recover from VIX shocks. Thus, the
transmission mechanism of shocks that affect VIX to commodity and financial markets
is more persistent than that of shocks to stock prices of U.S. technology and energy firms,
being primarily visible at horizons longer than a week. This greater persistence of shocks
to VIX could be related to the fact that for market participants elevated levels of implied
volatility in the stock market are indissolubly linked to strong corrections in asset markets
in the near future. On the contrary, WTI oil futures contracts, the U.S. default spread,
yields on U.S. 10-Treasuries and the perception of uncertainty in the U.S. Treasury market
appear as substantial receivers of net pairwise return and volatility connectedness in the
short- and long-term over most of the sample period.
Based on the magnitude of net pairwise connectedness measures, it is worth noting that
shocks to CESI have a particularly strong impact on crude oil prices mainly in the very
short-term and since the onset of the financial crisis. This result can be attributed to the
fact that crude oil prices are one of the major drivers of the equity market performance of
the fossil energy sector, so that current prices of conventional energy stocks generally
incorporate future prospects about crude oil prices. Furthermore, a significant net
pairwise volatility connectedness is observed between clean energy and high technology
stock prices, primarily in the short-term, over most of the period of study. This finding is
consistent with the notion that renewable energy stocks and technology stocks can be
considered as similar asset classes and is also in agreement with empirical evidence
reported by numerous previous studies (Bondia et al., 2016; Inchauspe et al., 2015; Kumar
et al., 2012; Sadorsky, 2012).
INSERT FIGURE 10 ABOUT HERE
INSERT FIGURE 11 ABOUT HERE
Finally, in this section we assess the robustness of our empirical results by examining the
sensitivity of the estimated spillover measures to the choice of the size of the rolling
window. To that end, Figure 12 plots the time-varying total return and volatility spillover
index estimates based on the conventional Diebold-Yilmaz framework for three
alternative rolling window lengths (150, 200 and 250 days), being 200 days the window

26
size considered in the baseline empirical analysis. Visual inspection of these graphs
clearly reveals that the estimates of time-varying total spillover indices remain
qualitatively and quantitatively unaffected by the chosen rolling window size, thus
validating the findings of our initial empirical analysis.10
INSERT FIGURE 12 ABOUT HERE
5.4. Discussion of results
Overall, the results of the time-frequency approach of Barunik and Krehlik (2018)
indicate that most of return and volatility connectedness among stock prices of renewable
and conventional energy firms and technology firms, WTI oil prices and the remaining
selected financial variables is generated at the higher frequency band (up to 5 days). This
evidence points to a very fast processing of information by commodity and financial
markets, so that both return and volatility shocks are essentially transmitted within one
week. Moreover, the increased degree of interconnectedness since the beginning of the
financial crisis implies that there is an evident risk of contagion across markets in the very
short-term, especially as a result of the general re-pricing of risk in the economy
propitiated by the global financial crisis. Two recent papers by Lau et al. (2017) and
Tiwari et al. (2018) have also documented the predominance of the short-term over the
long-term as the primary source of connectedness across white precious metal markets
and several financial asset classes, respectively, based on the Barunik-Krehlik method. In
turn, Krehlik and Barunik (2017) have shown the increasing importance of the effect of
volatility shocks up to one week in overall connectedness across oil-based commodity
markets also employing the Barunik-Krehlik framework. In contrast, the much lower
return and volatility connectedness at longer time horizons (more than a week) indicates
that financial and crude oil shocks are generally short-lived in such a way that the
development of commodity and financial markets in the long-term is mainly determined
by their own fundamentals.
The finding that crude oil price shocks have hardly any influence on the behavior of equity
prices of clean energy firms is also particularly important as it means a decoupling of the
renewable energy sector from the crude oil market in recent years. There are two key
reasons for the poor link between oil prices and alternative energy stocks. First, as argued
10 For the sake of brevity, the estimates of return and volatility connectedness measures at higher and shorter
frequency bands for different rolling window sizes are not displayed. However, the empirical results of the
alternative window lengths are not materially different from those presented in this section.

27
by Auran and Gullaksen (2017), the dynamics of energy markets has changed
dramatically in the twenty-first century as crude oil and renewable energy no longer
compete on the same markets. As a matter of fact, crude oil is basically used to produce
transportation fuels, while clean energy is principally utilized to generate electricity.
Therefore, by not being direct substitutes, when the price of one of them falls, the demand
for the other does not necessarily decreases.11 Second, the continual drop in capital,
operating and financing costs for alternative energy projects over the last years has
increased the margins captured by leading players in the green energy industry. Thus, the
increasing cost-competitiveness of renewables together with the improved technology for
alternative energy production may also have contributed to the decoupling of green
energy stocks from the evolution of crude oil prices in recent years. Various papers, such
as those of Ahmad (2017), Auran and Gullaksen (2017), Henriques and Sadorsky (2008)
and Sadorsky (2012), have also documented the weakening of the relationship between
crude oil prices and equity prices of new energy companies.
Furthermore, the role of crude oil prices as a major net receiver of financial shocks clearly
shows that oil price fluctuations are not exogenous, but they are a part of the global
financial system and, therefore, are significantly affected by shocks in financial markets.
This evidence is consistent with the increasing financialization of the oil market (Junttila
et al., 2018; Krehlik and Barunik, 2017; Zhang, 2017) through which crude oil prices
have become more correlated with financial assets since the early 2000s. Specifically, the
growing participation of big institutional investors, hedge funds and other financial
institutions in the crude oil market is likely to lead to faster reactions of oil prices to
shocks due to the exploitation of possible arbitrage opportunities arising from deviations
of prices caused by such shocks. In other words, these financial participants are more
willing to re-balance more often their positions in oil-related assets. Hence, the price of
crude oil is no longer determined solely by its supply and demand, but it is also crucially
affected by changes in financial conditions as a result of the increasing interest of
investors in diversification. Therefore, this financialization process can be one of the key
factors behind the greater relative importance of short-term connectedness between crude
oil and financial markets.
11 The only exception in this comparison are liquid transport biofuels, e.g. biodiesel and bioethanol, which
directly compete with crude oil products.

28
6. Concluding remarks
The unprecedented growth of the clean energy sector in the last decade, coupled with the
fact that renewables are expected to be the fastest growing energy source over the next
twenty years, have generated large interest to know the stock market performance of
green energy firms and its interactions with crude oil prices and several major financial
indicators. This paper examines the time and frequency dynamics of connectedness
among stock prices of U.S. alternative energy companies, crude oil prices and a number
of influential financial variables, namely high technology and conventional energy stock
prices, U.S. 10-year Treasury bond yields, the U.S. default spread and the volatility of
U.S. equity and Treasury bond markets. To that end, the novel connectedness
methodology proposed by Barunik and Krehlik (2018), which can be seen as an extension
to the time-frequency space of the spillover index approach of Diebold and Yilmaz
(2012), is used. In essence, the Barunik-Krehlik method provides a suitable framework
for measuring the dynamics of return and volatility connectedness over time and across
frequencies simultaneously.
Our empirical results reveal the overwhelming predominance of the higher frequency
band (up to 5 days) over the lower frequency band (more than 5 days) in terms of the
magnitude of return and volatility connectedness. This suggests that crude oil and
financial markets have become highly efficient and process information very quickly, so
that shock transmission effects occur primarily within one week. In turn, the lower long-
term connectedness is congruent with the widely accepted view that in the long run
markets tend to be primarily driven by their own fundamentals and the general economic
prospects. Furthermore, a greater degree of market linkage is observed since the outbreak
of the U.S. subprime mortgage crisis in summer of 2007, reflecting the global
reassessment and re-pricing of risk across the whole economy and the scenario of
increased interdependence caused by the recent worldwide financial crisis.
Another key finding is that crude oil prices do not play a critical role in explaining
movements in stock prices of clean energy companies in the short-term or the long-term
during the entire sample period. Thus, the equity market performance of renewable
energy firms is likely to be more related to factors such as technology innovation, capital
spending, legislation or geographic availability than to oil prices. Therefore, this result
supports the decoupling of the alternative energy sector from the crude oil market,
confirming that the dynamics of global energy markets has undergone a paradigm shift

29
over the last years. This transformation has been fostered, at least partly, by the fact that
crude oil and green energy are used to satisfy different parts of global energy demand,
along with the higher cost-competitiveness of renewables and the awareness about the
importance of clean energy in combating climate change. Accordingly, it can be stated
that the recent crude oil price development does not appear to be a crucial factor in the
ongoing transition to a low-carbon energy system.
The empirical evidence in this study may have important implications for various
economic agents in terms of portfolio construction and risk management at different
investment horizons. For short-term investors and portfolio managers, it is very difficult
to find interesting diversification opportunities across crude oil and financial markets in
the short-term, particularly during episodes of heightened financial turmoil, in their
attempt to reduce contagion risk. In this regard, agents with short investment horizons
should especially avoid clean energy and high technology stocks in a portfolio due to the
significant short-term volatility connectedness between both types of stocks. Moreover,
short-term investors could profit from the significant information regarding the
development of crude oil prices in the very short-term contained in stock prices of
conventional energy companies. In contrast, investors and portfolio managers with longer
horizons can benefit from sizeable hedging and diversification opportunities by including
crude oil-related financial products in portfolios consisting of alternative energy and/or
high technology stocks. Additionally, policy makers should be aware that the alternative
energy sector does not require short- and/or long-term specific policy measures of
protection against crude oil price shocks in order to facilitate the transition towards a
sustainable energy system. Instead, policies promoting renewable energy development
should be preferably oriented towards improving investment and green energy-related
technology.
References
Ahmad, W. (2017). On the dynamic dependence and investment performance of crude
oil and clean energy stocks. Research in International Business and Finance, 42, 376-
389.
Auran, I.K. and Gullaksen, J.F. (2017). Time-varying risk factor models for renewable
stocks. Master’s thesis Norwegian University of Science and Technology (NTNU).
Available at http://hdl.handle.net/11250/2468842
Bhardwaj, G., Gorton, G. and Rouwenhorst, G. (2015). Facts and Fantasies about
commodity futures ten years later. NBER Working Paper No. 21243.

30
Barunik, J. and Krehlik, T. (2018). Measuring the frequency dynamics of financial and
connectedness and systemic risk. Journal of Financial Econometrics, 16, 271-296.
Bondia, R., Ghosh, S. and Kanjilal, K. (2016). International crude oil prices and the stock
prices of clean energy and technology companies: Evidence from non-linear cointegration
tests with unknown structural breaks. Energy, 101, 558-565.
Boyer, M.M. and Filion, D. (2007). Common and fundamental factors in stock returns of
Canadian oil and gas companies. Energy Economics, 29, 428-453.
British Petroleum (2017). BP Statistical Review of World Energy 2017
Corbet, S., Meegan, A., Larkin, C.J., Lucey, B.M. and Yarovaya, L. (2018). Exploring
the dynamic relationships between cryptocurrencies and other financial assets. Economics
Letters, 165, 28-34.
Creti, A., Joëts, M. and Mignon, V. (2013). On the links between stock and commodity
markets’ volatility. Energy Economics, 37, 16-28.
Dew-Becker, I, and Giglio, S. (2016). Asset pricing in the frequency domain: Theory and
empirics. The Review of Financial Studies, 29, 2029-2068.
Diebold, F.X. and Yilmaz, K. (2012). Better to give than to receive: Predictive directional
measurement of volatility spillovers. International Journal of Forecasting, 28, 57-66.
Dutta, A. (2017). Oil price uncertainty and clean energy stock returns: New evidence
from crude oil volatility index. Journal of Cleaner Production, 164, 1157-1166.
Ferrer, R., Bolós, V.J. and Benítez, R. (2016). Interest rate changes and stock returns: A
European multi-country study with wavelets. International Review of Economics and
Finance, 44, 1-12.
Gilchrist, S. and Mojon, B. (2014). Credit risk in the euro area. NBER Working Paper
No. 20041.
Gilchrist, S. and Zakrajšek, E. (2012). Credit Spreads and Business Cycle Fluctuations.
American Economic Review, 102, 1692–1720.
Henriques, I. and Sadorsky, P. (2008). Oil prices and the stock prices of alternative energy
companies. Energy Economics, 30, 998-1010.
Inchauspe, J., Ripple, D.R. and Trück, S. (2015). The dynamics of returns on renewable
energy companies: A state-space approach. Energy Economics, 48, 325-335.
International Energy Agency (2017). World energy outlook 2017. (International Energy
Agency).
IRENA (2017). Renewable Energy Statistics 2017. Technical report, IRENA.
Jareño, F., Ferrer, R. and Miroslavova, S. (2016). US stock market sensitivity to interest
and inflation rates: A quantile regression approach. Applied Economics, 48, 2469-2481.
Junttila, J., Pesonen, J. and Raatikainen, J. (2018). Commodity market based hedging
against stock market risk in times of financial crisis: The case of crude oil and gold.
Journal of International Financial Markets, Institutions and Money,
Koop, G., Pesaran, M.H. and Potter, S.M. (1996). Impulse response analysis in nonlinear
multivariate models. Journal of Econometrics, 74, 119-147.

31
Krehlik, T. and Barunik, J. (2017). Cyclical properties of supply-side and demand-side
shocks in oil-based commodity markets. Energy Economics, 65, 208-218.
Kumar, S., Managi, S. and Matsuda, A. (2012). Stock prices of clean energy firms, oil
and carbon markets: A vector autoregressive analysis. Energy Economics, 34, 215-226.
Lau, M.C.K., Vigne, S.A., Wang, S. and Yarovaya, L. (2017). Return spillovers between
white precious metal ETFs: The role of oil, gold, and global equity. International Review
of Financial Analysis, 52, 316-332.
Li, L., Yin, L. and Zhou, Y. (2016). Exogenous shocks and the spillover effects between
uncertainty and oil price. Energy Economics, 54, 224-234.
Managi, S. and Okimoto, T. (2013). Does the price of oil interact with clean energy prices
in the stock market? Japan and the World Economy, 27, 1-9.
Moya-Martínez, P., Ferrer-Lapeña, R. and Escribano-Sotos, F. (2014). Oil Price risk in
the Spanish stock market: An industry perspective. Economic Modelling, 37, 280-290.
Moya-Martínez, P., Ferrer-Lapeña, R. and Escribano-Sotos, F. (2015). Interest rate
changes and stock returns in Spain: A wavelet analysis. BRQ Business Research
Quarterly, 18(2), 95-110.
Pesaran, H.H. and Shin, Y. (1998). Generalized impulse response analysis in linear
multivariate models. Economics Letters, 58, 17-29.
Reboredo, J.C. (2015). Is there dependence and systemic risk between oil and renewable
energy stock prices? Energy Economics, 48, 32-45.
Reboredo, J.C., Rivera-Castro, M.A. and Ugolini, A. (2017). Wavelet-based test of co-
movement and causality between oil and renewable energy stock prices. Energy
Economics, 61, 241-252.
Reilly, F.K., Wright, D.J. and Johnson, R.R. (2007). Analysis of the interest rate
sensitivity of common stocks. The Journal of Portfolio Management, 33, 85-107.
REN21 (2017). Renewables 2017 – Global Status Report. Technical report, REN21.
Sadorsky, P. (2001). Risk factors in stock returns of Canadian oil and gas companies.
Energy Economics, 23, 17-28.
Sadorsky, P. (2012). Correlations and volatility spillovers between oil prices and the stock
prices of clean energy and technology companies. Energy Economics, 34, 248-255.
Stiassny, A. (1996). A spectral decomposition for structural VAR models. Empirical
Economics, 21(4), 535-555.
Tiwari, A.K., Cuñado, J., Gupta, R. and Wohar, M.E. (2018). Volatility spillovers across
global asset classes: Evidence from time and frequency domains. The Quarterly Review
of Economics and Finance, In press. DOI: https://doi.org/10.1016/j.qref.2018.05.001.
Zhang, D. (2017). Oil shocks and stock markets revisited: Measuring connectedness from
a global perspective. Energy Economics, 62, 323-333.
Zhang, G. and Du, Z. (2017). Co-movements among the stock prices of new energy, high-
technology and fossil fuel companies in China, Energy, 135, 249-256.