Resources Policy 72 (2021) 102112

Contents lists available at ScienceDirect

Resources Policy
journal homepage: [Link]/locate/resourpol

Dynamic connectedness between uncertainty and energy markets: Do

investor sentiments matter?
Ata Assaf a, b, *, Husni Charif a, Khaled Mokni c, d
a
Faculty of Business and Management, University of Balamand, [Link]: 100, Tripoli, Lebanon
b
Cyprus International Institute of Management (CIIM), 2151, Nicosia, P. O. Box 20378, Cyprus
c
College of Business Administration, Northern Border University, Arar, 91431, Saudi Arabia
d
Institut Supérieur de Gestion de Gabès, Gabès University, Gabès, 6002, Tunisia

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

JEL classification: This paper examines the dynamics between the energy markets and uncertainty indices from January 1st, 2001
C54 till July 1st, 2020, using the time-domain TVP-VAR-based connectedness approach of Antonakakis and Gabauer
G10 (2017). In particular, we examine whether Economic Policy Uncertainty (EPU), Geopolitical Risk (GPR), World
Q40
Trade Uncertainty (WTU), and Equity Market Uncertainty (EMU) have an impact on the dynamics of returns of
Keywords: oil, gas, and coal markets. Results suggest that the average influence of market uncertainty on energy markets is
Energy prices
approximately 53%. Second, we find that the EPU contributes the most to the energy markets, followed by the
Uncertainty
TVP-VAR
World Trade Uncertainty index. Third, considering the energy markets, we find that the oil markets contribute
Dynamic connectedness the most to other markets. Fourth, we find that the EMU receives the most contribution from other markets,
Investor sentiment followed by the EPU. Finally, we observe that the total connectedness index is relatively high, coinciding with the
Global Financial Crisis of 2008 and the COVID-19 pandemic outbreak. Further analysis of the possible effect of
the investor sentiment on the dynamic connectedness shows that the consumer sentiment index (CSI) has a
negative (positive) effect on the uncertainty (energy) net connectedness. Our findings have important implica­
tions for risk management in energy markets and entail some policy implications for regulators.

1. Introduction policymakers, and authorities’ decision-driven, investor behavior is

influenced by these decisions. Thus, the lack of policymakers’ credibility
In recent decades, the use of energy products, including crude oil, and their unknown intentions can raise uncertainties that may affect
natural gas, and coal, in almost all economic activities has been investors’ decisions and impact energy as well as other commodities’
increased and occupied a pivotal position, further strengthened after the returns. In contrast, proper disclosure of information and authorities’
global financial crisis of 2008. In the same period, the economic policy good repute can ease down uncertainty. Thus, this necessitates that
uncertainty (EPU) is augmented remarkably due to the influence of policymakers be wary of the various uncertainty impacts in order to
economic and financial pressures, as well as the recent health crisis of regulate the economy accordingly, whereas investors need to take pre­
the COVID-19 pandemic, leading to an intensive fluctuation of energy ventive measures to reduce its adverse effects. (see Figs. 3–6)
prices. Therefore, the debate over the energy prices-EPU has been Due to the ever-increasing demand for energy commodities, crude
renewed to determine and understand the type and the strength of the oil, labeled as the king of commodity, as well as coal and natural gas,
relationship between these two phenomena. Accordingly, in the after­ play a major role in the global economy. In fact, the price volatility of
math of the 2008 Great Recession, analyzing the detrimental impact of these commodities plays a major factor in option pricing, portfolio
economic policy uncertainty (EPU) and other risk indices on financial allocation, and hedging strategies (Antonakakis et al., 2018). Thus, ac­
and various energy markets have been at the epicenter of academic curate modeling of their features and characteristics promotes the cap­
research interests, which focused on building an understanding of the ital market’s health (Liang, 2020). This justifies the richness of recent
transmission mechanisms between them. Wang and Lee (2020) argue studies analyzing the spillover effects and connectedness between crude
that since most global energy investments are government-driven, oil prices and other energy commodities with external environmental

* Corresponding author. Faculty of Business and Management, University of Balamand, [Link]: 100, Tripoli, Lebanon.
E-mail addresses: [Link]@[Link] (A. Assaf), [Link]@[Link] (H. Charif), kmokni@[Link] (K. Mokni).

[Link]
Received 6 December 2020; Received in revised form 25 February 2021; Accepted 8 April 2021
Available online 4 May 2021
0301-4207/© 2021 Elsevier Ltd. All rights reserved.
A. Assaf et al. Resources Policy 72 (2021) 102112

factors that relate to risk and uncertainty of which we opt to use the the year 2020. Finally, a robustness analysis dealing with an individual
text-based macroeconomic uncertainty indices proposed by Baker et al. analysis between each uncertainty measure and energy prices confirms
(2016). our findings. It provides a more detailed and comprehensive picture of
The main goal of this study is to investigate the spillover effects the energy-uncertainty nexus. Finally, the results generally report a
between US economic policy uncertainty (EPU), geopolitical risk (GPR), significant impact of investor sentiment on the dynamic connectedness
US equity market uncertainty (EMU), and the world trade uncertainty between uncertainty indices and energy markets. More precisely, we
index (WTU) with various energy market indices, including oil, gas, and find that the total connectedness is negatively affected by the consumer
coal. We employ the time-domain TVP-VAR-based connectedness sentiment index (CSI). However, the CSI has a negative (positive) effect
approach of Antonakakis and Gabauer (2017), allowing for time-varying on the uncertainty (energy) net connectedness. These findings are useful
effects of uncertainty shocks on commodity prices (e.g., Huang et al. and have important implications for a variety of economic players
2021). We also consider the impact of energy prices on market un­ regarding policymakers, traders as well as financial investors.
certainties. For example, since oil prices are often considered endoge­ The remainder of this paper is organized as follows: Section 2 pre­
nous in the oil-economic growth relationship and natural gas may sents the literature review of how EPU, GPR, WTU, and EMU indices
impact the U.S. and world economies, we also investigate the impact of affect the energy market represented by indices for Oil, Gas, and Coal in
energy market prices on the different market uncertainties. To our particular. Section 3 illustrates the adopted methodology, while Section
knowledge, our work is the first to analyze the combined dynamical 4 describes the data and empirical results. Section 5 concludes.
effects of these measures on energy markets.
Our study seeks to make some contributions to the literature on 2. Literature review
several fronts. Firstly, we move the debate on the economic uncertainty
– energy price nexus to the different types of uncertainty. In fact, our Although energy commodity prices are usually determined in global
paper is the first to consider the different categories of uncertainty, markets by the law of demand and supply, research has identified the
including EPU, EMU, GPR, and WTU, with the prices of different energy presence of a dynamic interconnectedness with various risk and uncer­
products (oil, natural gas, and coal). Therefore, we extend previous tainty indices (Guo et al., 2018; Bouiyour et al., 2019; Liu et al., 2019;
studies focusing only on the EPU in relation to the energy market (Arouri Demirer et al., 2019). It is known that energy markets are characterized
et al., 2014; Wei et al., 2017; You et al., 2017; Bahmani et al., 2018; by scarcity, with a low price elasticity of demand and energy prices
Scarcioffolo and Etienne, 2018; Ma et al., 2019; Yang, 2019; Haile­ being highly vulnerable to geopolitical risks (Su et al. (2019)). Geopo­
mariam et al., 2019 Chen, J. et al., 2019; Chen, X. et al. 2019) and others litical risks are those risks related to terrorist acts, tensions, and wars
focusing only on oil prices to represent energy markets (Liang et al., between countries, as defined by Caldara and Iacoviello (2018). These
2020) by considering further categories of uncertainty and energy pri­ extreme risks greatly impact the energy markets, as reflected in future
ces. Second, our analysis is based on the sophisticated TVP-VAR expectations about future energy supply and demand (Peng et al.
approach having the advantage to overcome the weakness of the (2018); Bouoiyour et al. (2019); Liu et al. (2019); Demirer et al. (2019)).
rolling-window approach, which lead to erratic or flattened parameters The extreme geopolitical events may cause panic among investors,
and a loss of valuable observations (Antonakakis and Gabauer, 2017; leading to abnormal market fluctuations and then affecting energy
Antonakakis et al., 2018, 2020; Korobilis and Yilmaz, 2018). Third, our markets returns and volatility (Antonakakis et al. (2017a); Plakandaras
analysis considers the energy prices in different key regions to investi­ et al. (2019); Mei et al. (2020)). On the other hand, the adverse effects of
gate whether the relationship between energy and uncertainty is uncertainty on the macroeconomy are commonly attributed to the the­
affected by geographic factors. It is the first study that considers this ory of irreversible investment under uncertainty (e.g., see Cristou et al.
factor regarding the energy-uncertainty nexus to the best of our (2020)).
knowledge. Third, a further analysis considered by this study attempts to Authors have implemented various modeling techniques to capture
explain the dynamic connectedness between uncertainty indices and transmission mechanisms between indices of risk and uncertainty and
energy prices by the investor sentiment measured by the consumer energy market commodities, including crude oil, natural gas, and coal
sentiment index (CSI). In fact, the literature argues the existence of markets (e.g., Aloui et al. (2016), Yin (2016), Feng et al. (2020), and
significant linkage between energy markets and sentiment on the one Huang et al. (2021)). Studies investigated the impact of different un­
hand (Maslyuk-Escobedo et al., 2017; Qadan and Nama, 2018) and certainty indices using different methods, such as structural VAR models
between uncertainty and investors’ sentiment on the other hand (Zhang, (i.e., Kang and Ratti (2013); Antonakakis et al. (2014)), copula approach
2019; Rehman and Apergis, 2019; Nartea et al., 2020). This linkage (i.e., Aloui et al. (2016)), GARCH, and GARCH-MIDAS models (i.e., Wei
motivates one to investigate the role of the investors’ sentiment in et al. (2017); Ma et al. (2019); Scarcioffolo and Etienne (2018)),
driving the dynamic connectedness between uncertainty and energy Nonlinear Distributed Lag (NARDL) models (Bahmani et al. (2018)), and
prices. time-varying factor-augmented models (TVP-FAVAR), among others
Our contribution falls along with the following main findings. First, (Feng et al. (2020); and Huang et al. (2021)).
the average influence of market uncertainty on energy markets is A rich strand of the literature confirmed the effects of uncertainty in
approximately 53%. Second, we find that the EPU, followed by the WTU, economic policy, equity markets and trade, and profound geopolitical
contributes the most to the energy markets. For example, EPU and WTU risk on the energy market. Following the seminal work of Baker et al.
contribute 57.03% and 30.12% each to other markets. Third, consid­ (2016, 2019b), many scholars studied the link between the uncertainty
ering which energy sector contributes the most to other markets, the oil indices (EPU, EMU) and the oil market or its components. Since eco­
market, on average, contributes 83%–90% to the forecast error variance nomic policy drives several macroeconomic indices, any uncertainty in
of all other markets, followed by the coal energy sector. Fourth, these policies may affect the investors’ sentiments and slow down the
considering how the shocks coming from others affect those markets, we economy. Kang and Ratti (2013) reported that an increase in the US
find that EMU receives the most contribution from other markets by economic policy uncertainty is associated with oil demand shocks,
52%, then followed the Economic Policy Uncertainty by 42%. For the thereby connecting structural oil price shocks to economic policy un­
energy sector, the oil market receives about 70% contribution of shocks certainty. This work was extended by Antonakakis et al. (2014), who
coming from other markets, followed by the coal markets. Finally, we studied the dynamic spillovers of oil price shocks and EPU for a sample
observe that the total connectedness index is relatively high during the of both net oil-exporting and net oil-importing countries and found a
entire period, exceeding the 60% level at a point that coincides with the negative response in either way. They found that the EPU was the
Global Financial Crisis of 2008. Then, the TCI reached a level of 51%, dominant net transmitter of shocks between 1997 and 2009. Ajmi et al.
coinciding with the COVID-19 pandemic outbreak in the first quarter of (2014) investigated causality between EPU’s of several countries

2
A. Assaf et al. Resources Policy 72 (2021) 102112

representing major economies. Their results indicated the existence of uncertainties in the crude oil market.
bidirectional causality between countries’ economic policy uncertainty Similarly, using the oil volatility index (OVX), Gong and Lin (2018)
throughout the sample. Aloui et al. (2016) revealed the dynamic investigated whether investor fear gauge (IFG) contains incremental
dependence structure between equity and economic policy uncertainty information content for forecasting the volatility of crude oil futures.
and crude-oil returns. They reported that higher uncertainty signifi­ They employed heterogeneous autoregressive (HAR) models, incorpo­
cantly increases crude-oil returns in the period prior to the financial rating the investor fear gauge index. They provided evidence that the
crisis and Great Recession. Yin (2016) examined the time-varying cor­ in-sample and out-of-sample performances of HAR models with IFG are
relation and the mean and volatility spillover between the oil market significantly better than their corresponding HAR models without IFG in
and the EPU. They inferred the importance of policy uncertainty and predicting oil futures volatility. Then, Gong and Lin (2019) incorporated
provided evidence that both oil supply shocks and real economic shocks the leverage effects and structural changes in the HAR model to study
cause fluctuations over time in the correlation between the EPU and oil their effects on oil futures markets. Employing different versions of the
returns. HAR model, their results show that leverage effects and structural
Bahmani et al. (2018) studied the impact of country-specific policy changes contain significant information for predicting oil volatility, with
uncertainty on oil prices. They found short-run effects of policy uncer­ structural changes having more in-sample and out-of-sample incre­
tainty measures of several countries and long-run asymmetric effects mental information than leverage effect. However, leverage effects have
only in the case of China, which, in their opinion, reflects the recent more out-of-sample information for predicting 1-day volatility.
surge in China’s engagement in world trade. Wen et al. (2019) indicated As pointed out earlier, markets fluctuate in resonance with major
a predictive ability of EMU in the crude oil futures market. Similarly, events such as violence and political unrest, terrorist attacks, military-
Balcilar et al. (2017) reported that EPU and EMU have strong asym­ related, nuclear, or other adverse geopolitical tensions or threats (See
metric predictive power for oil returns, over virtually the entire distri­ Antonakakis et al. (2017a) and references therein). Such geopolitical
bution, with some exceptions in the tails. More recently, Feng et al. friction induces high-risk sentiments amongst investors, which
(2020) studied the time-varying impact of EPU on crude oil price fluc­ adversely affects global markets’ returns, volatility, and the covariances
tuations and explored the differences in this impact between net-oil amongst its constituents. In particular, the global energy supply has
exporting and net-oil importing country’s EPU. Findings indicate that been affected by the geopolitical turmoil in the currently turbulent
the impact of EPU on oil price fluctuations in these countries is het­ Middle East region, which contains a significant proportion of global oil
erogeneous due to the differences in their economic policies, fiscal reserves.1 This shortage impacts energy commodity prices and their
dependence, and risk preference. Liang et al. (2020) investigated the five volatility and affects the financial markets and economic output glob­
economic uncertainty indices predictability for oil price volatility. ally. Empirical evidence has shown that economic performance reacts
Empirical results suggest that global economic policy uncertainty GEPU negatively to major security risk-generating incidents around the globe.
and EMU indices have significant predictive power for crude oil market In order to capture upsurges in geopolitical tensions, studies used the
volatility. Wang and Lee (2020) studied the dynamic connectedness and news-based computer-generated geopolitical risk (GPR) index proposed
spillover effects between categorical policy uncertainty and WTI crude by Caldara and Iacoviello (2018) to study the effect of these geopolitical
oil returns for oil-importing countries and concluded that economic rigidities on the oil markets. For instance, based on a long historical
policy uncertainty is a net transmitter of the system. This coincides with monthly stock and oil data that spans over a century, Antonakakis et al.
results obtained by Yang (2019). Huang et al. (2021) analyzed the (2017a) studied whether the time-varying stock–oil association is
time-varying relationships between commodity prices and several un­ affected by geopolitical risk. The results reveal that geopolitical risk
certainty measures, including the EPU. Results confirmed that the effects triggers a negative effect, mainly on oil returns and volatility. Findings
of uncertainty shocks on energy commodity prices are generally revealed that the oil market-return means and variability are signifi­
time-varying. cantly affected by GPR. Demirer et al. (2019) concluded that the impact
Other studies focused on the effects on oil market volatility. Wei et al. of GPRs is primarily on the volatility of oil markets. Liu et al. (2019)
(2017) demonstrated that the EPU indices have significant and analyzed the role of geopolitical risks, especially serious geopolitical risk
comprehensive predictive power for oil price volatility instead of other (GPRS), in predicting oil volatility and reported that serious geopolitical
potential determinants. They concluded that the EPU indices are key risks (GPRs) significantly affect future oil volatility. Besides, Qin et al.
factors to consider when determining crude oil market volatility. Scar­ (2020) revealed that the effect of the GPRs varies according to the en­
cioffolo et al. (2018) concluded the existence of a significant own- and ergy market condition and the type of geopolitical (threats or acts) risks.
cross-market volatility spillover effect between the three variables. They In the same context, the empirical results of Mei (2020) suggested a
also found that innovations in EPU negatively affect the volatility of oil positive association between oil volatility and GPR uncertainty.
and natural gas returns, whereas the innovations in the two energy Other scholars applied models with time-varying parameters in order
markets positively affect the volatility of the economic policy uncer­ to capture the dynamics of the relationships as well. Plakandaras et al.
tainty. Ma et al. (2019) investigated the impact of EPU on crude oil (2019) found that war-related geopolitical risks could predict oil returns
returns volatility. Their study reported a positive and significant impact correctly in the short run. Li et al. (2020) studied the links between
of EPU on crude oil volatility. Liang et al. (2020) investigated the ability crude oil prices and geopolitical risks. Their results suggested the exis­
of several economic uncertainty indices, including the global EPU and tence of a co-movement between the two variables at high frequencies,
EMU, in predicting oil price volatility. Their results indicated that EMU while in the long run, there existed no links between the two variables.
has a better predictive ability for the realized variance of WTI crude oil. Su et al. (2019) studied the causal links between oil prices, geopolitical
It is also noted that the price volatility of commodities is closely risks, and financial liquidity. They indicated that when geopolitical risks
related to energy markets. For instance, Gong and Lin (2017) used the were high, oil prices were associated with financial liquidity in the time
heterogeneous autoregressive type (HAR-type) models to forecast the domain. Bouiyour et al. (2019) explored the response of the oil market to
good and bad uncertainties of crude oil prices. They investigated the geopolitical risks, with the results indicating a significant positive
effects of lagged bad and good uncertainties, daily positive and negative impact, while the effect of geopolitical threats was non-significant.
signed jump variations, and leverages in predicting good and bad un­ Cunado et al. (2019) studied the dynamic impact of geopolitical risks
certainties. The results indicate that lagged bad (or good) uncertainties, (GPRs) on real oil returns and reported that, in general, the former has a
daily positive signed jump variation, and daily negative signed jump
variation contain incremental out-of-sample information for forecasting
good (or bad) uncertainties. Their results also indicate that lagged le­ 1
64.5% of the OPEC total according to the OPEC Annual Statistical Bulletin
verages play an important role in forecasting the good and bad 2019.

3
A. Assaf et al. Resources Policy 72 (2021) 102112

significant negative impact on oil returns. However, Smales (2019) effect of the different market uncertainties on the oil, gas, and coal
concluded that geopolitical risk is associated with positive oil returns. Li markets.
et al. (2020a) found a significant dynamic co-movement and causality in
the time-frequency domains between crude oil prices and geopolitical 3. Empirical methodology
factors. The dynamic correlation is particularly reported to be strong
and has increased volatility during periods of political tensions. Results This section presents the econometric methodology used in the
unveiled unidirectional causality running from geopolitical factors to empirical analysis of the total and directional connectedness between
crude oil prices. Li et al. (2020b) investigated the frequency- and market uncertainties and energy markets. We first describe the meth­
time-varying co-movement and causal relationship between crude oil odology proposed by Diebold and Yilmaz (2009, 2012, 2014) and then
prices and geopolitical factors. Results demonstrated a high degree of outline the dynamic connectedness procedure based on TVP-VAR pro­
co-movement between geopolitical risks and oil prices in the short run vided by Antonakakis and Gabauer (2017). This approach seems to have
but not in the long run for most of the sample period. gained popularity, thus ensuing many studies (Korobilis, 2018; Yilmaz
Other studies focused on the effects of geopolitical risks on energy Gabauer and Gupta, 2018; Liu and Gong, 2020, Youssef et al., 2021,
market volatility. For example, Wang and Yang (2018) studied the in­ among others).
fluence of geopolitical risks on oil price volatility during the shale rev­
olution. They found that the response of oil price volatility to 3.1. TVP-VAR approach
geopolitical risks was higher with a shale production shock. Using the
GARCH-MIDAS model, Liu et al. (2019) examined the ability of Researchers have developed elaborate methods aiming to capture
geopolitical risks to predict oil price volatility, with the findings that shock transmission mechanisms between macro-economic variables.
geopolitical risks caused fluctuations in the oil market. Using a similar Diebold and Yılmaz (2009, 2012, 2014) (DY) introduced a
methodology, the Mixed Data Sampling (MIDAS) model, Mei et al. rolling-window VAR-based empirical approach providing various
(2020) studied the role of geopolitical risk uncertainty in forecasting oil connectedness measures built from pieces of variance decompositions,
futures market volatility. Their results indicated that geopolitical risk which they describe as “natural and insightful.” Their approach assesses
uncertainty positively affected oil realized volatility. Similarly, Asai forecast error variation in the variable i attributed to innovations in
et al. (2020) found that geopolitical risks could help in forecasting crude other variables in the model. They considered that the time series fol­
oil futures volatility. lows the reduced-form autoregressive (VAR) model with a
Another source of uncertainty that may impact energy commodities fixed-parameter and fixed variance-covariance matrix.
is the recent sharp rise in uncertainty due to the recent US trade policies Instead, in order to improve the accuracy of the dynamic connect­
and President Trump’s trade rhetoric. This includes the withdrawal from edness measures of DY, Antonakakis and Gabauer (2017) employed a
the Trans-Pacific Partnership, and the resurgence of “protectionism” time-varying parameter vector autoregressive model (TVP-VAR) with a
through tariff impositions by the Trump administration targeting China time-varying covariance structure proposed in Primiceri (2005), as
mostly, and consequent retaliation, which has escalated tensions with opposed to the constant-parameter rolling-window VAR approach. The
China to the brink of a trade war (Baker et al. (2019a); Berthou et al. drifting coefficients and stochastic volatility are meant to capture the
(2019)). Trade uncertainty is also aggravated by the foreseen reper­ possible nonlinearities or time variation in the lag structure of the
cussion of Brexit (Steinberg (2019)) and serious calls for the United model, the possible heteroscedasticity of the shocks, and the non­
States to withdraw from the World Trade Organization, as well as the linearities in the simultaneous relations among the variables of the
surging pandemic outbreak, to name a few. This uncertainty about the model (Primiceri, 2005). The TVP-VAR model has been used in many
outlook for trade policy worries policymakers as it may have long-term earlier studies but not in the context of connectedness. The TVP-VAR
negative implications on trade and investment (Caldara et al. (2020); dynamic connectedness approach improves the Diebold and Yilmaz
Handley and Limão (2017)). In fact, it reduced business confidence to methodology in many ways. It is argued by Antonakakis and Gabauer
historic lows (Kusek et al. (2020)), which had its effect on economic (2017), Antonakakis et al. (2020) that this framework allows “to capture
activity (Olasehinde-Williams (2020)), thus posing a threat to global possible changes in the underlying structure of the data in a more flex­
economic development. Besides, this uncertainty roiled the stock mar­ ible and robust manner” and that results indicate that the TVP-VAR
kets and caused an increase in risk premia (Ikonen et al. (2019)). It had a estimations are superior to those generated by rolling-windows. First,
significant negative causal relationship with the returns of Bitcoin since the rolling window size is not set arbitrarily, there is no loss of
(Gozgor et al. (2019)). Thus, analyzing the impact of WTU on energy information in calculating the dynamic measures of connectedness.
commodity prices is important in the midst of these threats. To this end, Second, since it is based on a multivariate Kalman filter, it is less sen­
Karabulut et al. (2020) studied the relationship and co-movements be­ sitive to the presence of outliers and thus adjusts immediately to events
tween commodity prices and world trade uncertainty, as measured by (e.g., Antonakakis et al. (2018); and Gabauer and Gupta (2018)).
the index developed by Ahir et al. (2019). Findings indicated positively The N-variable TVP-VAR(p) model can thus be written as follows,
correlated co-movements corresponding to important worldwide polit­
ical and economic events. Qin et al. (2020) highlighted the time-varying Yt = βt Zt− 1 + εt , εt | Ωt− 1 ∼ N(0, Σ t ) (1)
interactions between oil price (WTI) and several uncertainty indices,
βt = βt− 1 + ϑt , ϑt | Ωt− ∼ N(0, Rt ) (2)
including trade EPU, using the wavelet analysis based on the time and 1

frequency domains.
where Yt = (Y1t , Y2t , …YNt )’is a vector with N variables and Zt− 1 is an
Based on the above, it appears that many authors have implemented
Np × 1 conditional vector formed of the past p lags, p being the optimal
various methods to capture the transmission mechanisms between ( )′
indices of uncertainty and energy markets. Yet, most of the studies lag length with Zt− 1 = Yt− 1 , Yt− 2 , ⋯, Yt− p , and Ωt− 1 represents
investigated the impact of a single index on these markets. In this paper, all information available through time t − 1, whereas βt is an N × Np
we particularly study the interconnectedness between oil, gas, and coal dimensional time-varying coefficient matrix, which follows a random
( )
energy markets on one side and the uncertainty indices on another. walk model and can be represented as βt = β1t , β2t , ⋯, βpt , βit
Specifically, we consider four market uncertainty indices: Geopolitical being an N × N matrix of time-varying coefficients.
risk, Economic Policy Uncertainty, World Trade Uncertainty, and Equity The error-disturbance vectors εt and ϑt are N × 1 and N ×
Market Uncertainty. We then employ the time-varying vector autore­ Np dimensional with N × N and Np × Np time-varying variance-covari­
gressive (TV-VAR) model to study their impact on the energy markets. ance matrices, Σt and Rt respectively.
Our results should be of particular interest to researchers studying the The Kalman filter algorithm is employed with forgetting factors

4
A. Assaf et al. Resources Policy 72 (2021) 102112

chosen based on a Bayesian model selection, as introduced by Koop and Coal Australia, Coal Columbia, and Coal South Africa. Then for the gas
Korobilis (2014) and demonstrated in Antonakakis et al. (2018). sector, we choose series for the US, Europe, and Japan. Also, we include
In order to calculate the generalized forecast error variance de­ an overall index for the gas market. All the data was sourced from the
compositions (GFEVD), the model in equation (1) is transformed to its World Bank Commodity Price Data available at [Link]
vector moving average (VMA) representation based on the Wold rep­ [Link]/en/research/commodity-markets updated on July 2020.
resentation theorem. The representation of the system is: We use three series for the oil markets, namely Brent Oil, Dubai Oil,
and WTI. Brent crude oil is extracted from the North Sea oil field, and the
∑
∞
Yt = Θjt εt− j (3) price is determined by the oil market situation in Europe and Africa.
j=0 However, WTI crude oil is extracted from the US oil field and is closer to
the US economy. Brent oil prices are the benchmark for crude oil in
where Θjt is an N × N dimensional matrix. Africa, Europe, and the Middle East. The pricing mechanism of Brent
crude oil determines about two-thirds of the world’s crude oil produc­
3.2. Dynamic connectedness measures tion. Its pricing takes into account geopolitical risk factors (Liang, 2020).
Data on the uncertainty indices, namely the Economic Policy Un­
To measure the dynamic connectedness between different variables, certainty (EPU), the World Trade Uncertainty (WTU), and Equity Market
the time-varying parameters and variance-covariance matrices of the Uncertainty Index (EMU), is obtained from [Link]
TVP-VAR model are used in Diebold and Yilmaz’s measure of connect­ m. EPU indices are developed for several countries worldwide. Among
edness. The elements of the dynamic H-step generalized variance the different measures of the latent “policy uncertainty” factor, the EPU
gH gH index developed by Baker et al. (2016) has become a benchmark for
decomposition matrix Dt = [dij,t ] are defined as:
measuring policy-related economic uncertainty in the United States due
∑H− 1 ( ′ )2 to its ease of use and wider coverage (André et al., 2017). The index, first
σ−jj,t1 h=0 ei Θh,t Σt ej
gH
dij,t =∑ ( ) , proposed in 1985, can measure the uncertainty in monetary, fiscal, and
other relevant policies. Results reported in Baker et al. (2016) showed a
H− 1 ′ ′
h=0 ei Θh,t Σt Θh,t ej
significant negative relationship between EPU and macro variables and
where σ jj,t is the jth diagonal element of Σt . The normalized terms ̃
dij,t =
gH can even explain the large fluctuation in the equity market2.
gH
For the Geopolitical Risk index, we rely on Caldara and Iacoviello
d
∑Nij,t gH
are used to determine the dynamic total directional connectedness, (2018), who constructed monthly indices of GPRs by counting the
d
j=1 ij,t
occurrence of words related to geopolitical tensions in leading interna­
net total directional connectedness, and total connectedness as follows. tional newspapers. In constructing the GPR index, they search for arti­
The total connectedness index (TCI), which measures interconnec­ cles containing references to words associated with: explicit mentions of
tedness among the time-series, is calculated as: geopolitical risk, as well as mentions of military-related tensions
∑N ̃gH involving large regions of the world and a U.S. involvement; nuclear
=j d ij,t
CtgH =
i, j=1, i∕
∑N gH × 100 (4) tensions; war threats and terrorist threats; actual adverse geopolitical
̃
j=1 d ij,t events (as opposed to just risks) which can be reasonably expected to
lead to increases in geopolitical uncertainty, such as terrorist acts or the
The directional volatility spillover received by variable i from all other
beginning of a war. The data is obtained from their website.3
variables j, is measured as:
As also the recent developments such as Trump’s trade war with
∑N ̃gH China and Brexit might be potential sources of uncertainty, we consider
=j d ij,t
(5) another uncertainty measure related to the World Trade Uncertainty. In
gH j=1, i∕
Ci←j = ∑ gH × 100
N ̃
i=1 d ij,t this paper, we use the World Trade Uncertainty Index (henceforth WTU)
Likewise, the spillovers received by variable j from all other variables developed by Ahir et al. (2019), which captures all the potential sources
i, is measured as: of uncertainty during the last two decades.4.
Table 1 reports the descriptive statistics on the log-returns of the
∑N ̃gH
=j d ij,t
energy market indices selected in this study and uncertainty market
(6)
gH j=1, i∕
Ci→j = ∑
N ̃gH
× 100 indices. We observe that Dubai Oil, Coal Columbia, and Natural Gas
j=1 d ij,t
Japan show the highest mean returns, while the indices for Natural Gas
In order to obtain the net pairwise directional connectedness, we sub­ of US, Europe, and the overall index have a negative mean rate of return.
tract the total directional connectedness to others from total directional Measuring risk by the variance, the table shows that US natural gas
connectedness from others, which can be interpreted as the influencing shows the highest risk with a variance of 170.69, followed by the oil
variable i has on the analyzed network. That is, markets, showing the riskiest among the group of the considered energy
markets. Whereas the European and Japanese gas markets present the
gH
Cij,t gH
= Cj←i,t gH
− Ci←j,t . (7) lowest risk levels. All the considered oil series are skewed to the left, as
indicated by the significant negative values of the skewness. While the
At last, the net pairwise directional connectedness is defined as:
gH gH gH coal markets for Columbia and South Africa are skewed to the right,
NPDCij = (̃ dji,t − ̃
dij,t ) × 100. A result greater than zero indicates
that variable i dominates variable j; otherwise, the latter is said to
dominate.
2
Studies that have focused on the effect of EPU on financial markets include
4. Data and empirical results those of Ajmi et al., (2015), Anoruo, Akpom, & Nwoye, (2017); Antonakakis,
Chatziantoniou, & Filis, (2013); Antonakakis, Chatziantoniou, & Filis, (2014);
Arouri, Estay, Rault, & Roubaud, (2016); Aye (2018); Aye, G. C., Clance, M. W.,
4.1. Data analysis
& Gupta, R. (2019); Bekiros, Gupta, & Majumdar, (2016), Bianconi et al.,
(2019); Bloom, (2009); Brogaard & Detzel (2015); Li, Balcilar, Gupta, & Chang,
We collect data for the monthly historical records between January (2016).
2001 and July 2020 on the energy sectors, including oil, coal, and nat­ 3
The indices can be found at: [Link]
ural gas sectors. We use three mostly used data series for the oil sector, htm.
namely Brent Oil, Dubai Oil, and WTI. For the coal sector, we school 4
The index can be retrieved from: [Link]

5
A. Assaf et al. Resources Policy 72 (2021) 102112

Table 1
Summary Statistics of the data series.
Mean Variance Skewness Kurtosis JB ERS Q(20) Q2(20) LM(20)

EPU 125.60 2279.03 0.906*** (0.000) 0.661* 34.39*** − 2.133** 29.76*** 42.45*** 12.80
(0.064) (0.000) (0.034) (0.000) (0.000) (0.242)
GPR 120.19 4045.05 2.033*** 5.95*** 481.05*** (0.000) − 1.31 (0.190) 334.31*** 87.0*** 23.40***
(0.000) (0.000) (0.000) (0.000) (0.000)
WTU 100.62 27468 4.04*** 20.33*** 4430*** 2.15 (0.03) 495.03*** 11.64 (0.341) 31.07***
(0.000) (0.000) (0.000) (0.000) (0.000)
EMU 20.18 66.38 2.63*** 10.10*** 1201.6*** − 1.632 203.78*** 23.46*** 5.22
(0.000) (0.000) (0.000) (0.10) (0.000) (0.000) (0.94)
Oil (Brent) 0.394 77.49 − 0.919*** 1.21*** 44.88*** − 2.28** 28.07*** 88.18*** 7.932
(0.000) (0.000) (0.000) (0.024) (0.000) (0.000) (0.727)
Oil (Dubai) 0.467 72.50 − 1.06*** 2.030*** 79.83*** − 1.99* 41.09*** 60.93*** 12.51
(0.000) (0.000) (0.000) (0.061) (0.000) (0.000) (0.263)
Oil (WTI) 0.34 74.82 − 0.84*** 1.50*** 47.42*** − 4.61*** 37.04*** 88.33*** 12.21
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.28)
Coal (Australia) 0.23 65.35 − 1.06*** 9.42*** 863.54*** − 4.26*** 25.09*** 0.823 12.76
(0.000) (0.000) (0.000) (0.000) (0.000) (1.00) (0.245)
Coal (Colombia) 0.343 50.146 0.108 2.13*** 42.35*** − 1.78* 77.86*** 18.86** 24.79***
(0.50) (0.000) (0.000) (0.07) (0.000) (0.028) (0.000)
Coal (S Africa) 0.285 47.37 0.038 1.82*** 30.79*** − 3.57*** 63.73*** 16.56* 21.61***
(0.811) (0.000) (0.000) (0.000) (0.000) (0.069) (0.008)
Nat Gas (US) − 0.563 170.69 0.176 0.82** 7.367** − 0.452 10.273 3.85 8.254
(0.272) (0.032) (0.025) (0.652) (0.476) (0.088) (0.692)
Nat Gas (Europe) − 0.113 39.10 − 0.814*** 2.37*** 76.53*** − 3.215*** 61.84*** 34.35*** 34.47***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.0000 (0.000)
Nat Gas (Japan) 0.329 22.98 − 0.974*** 2.513*** 93.53*** − 2.84*** 93.54*** 31.61*** 18.48**
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.031)
Nat Gas (Index) − 0.389 78.6 0.095 1.29*** 15.94*** − 0.486 16.26* 5.44 10.65
(0.554) (0.000) (0.000) (0.628) (0.077) (0.935) (0.431)

Notes: This table presents the descriptive statistics of the considered energy markets and policy uncertainty indices. J.B. is the Jarque-Bera normality test statistic. ERP
is the Elliot-Rothenberg- Stock unit root tests. Q(10) and Q2(10) are the Ljung-Box tests for 20th order serial correlations for returns and squared returns, respectively.
L.M. (20) is the Langrange Multiplier heteroscedasticity test at order 20. (***), (**), and (*) indicate the statistical significance, respectively, at the 1%, 5%, and 10%
levels.

their skewness is not significant. Most series also are characterized by implementation of quantitative easing stimulus packages to end the
excess kurtosis, indicating a leptokurtic distribution with fat tails. The recession. Recently, over the first quarter of 2020, the EPU has been
normality hypothesis is rejected for all series using the Jarque-Bera test rising due to the debt ceiling, COVID-19, and other political turmoil
at the 1% significance level. occurring across the world. As stated by Baker et al. (2016), the average
To test for stationarity in the returns series, we depend on the Elliot- economic policy uncertainty has doubled since 2008.
Rothenberg-Stock (ERS) test. The results are also reported in Table 1. For the energy markets, we can see large fluctuations from 2002 on.
The results show that all series are integrated of order zero. In testing The volatility in energy markets may be due to the geopolitics issues
whether series are autocorrelated or not, we rely on the LM tests. The such as Middle East tension and the increasing demand for oil from
results suggest that all series are autocorrelated, except for the case of emerging countries such as China and India. Natural gas prices were also
the US Natural Gas. affected by high oil prices and natural disasters like hurricanes. Their
As a preliminary exercise in the energy–certainty relationship, we prices peaked in 2008 before the Global Financial Crisis by then dropped
present the correlation matrix among the different variables in Table 2. after that. Unlike oil prices, natural gas prices have been lower since
We notice a negative correlation between the uncertainty indices and 2008. This can be attributed to the popularization of horizontal drilling
the rest of the energy markets. Regardless of the index we use, all the and hydraulic fracturing techniques that made natural gas production
correlation coefficients are negative, except for the case of the Word from shale formation economically profitable. Due to the production of
Trade Uncertainty index and the Natural gas index, where the correla­ shale oil in the United States and oversupply from major oil-exporting
tion is positive. This finding implies that an increase in the uncertainty countries, oil prices have been kept at a low level since 2015, only to
level leads to a decrease in energy markets’ returns. Another important rise slightly toward the end of our sample.
observation from Table 2 is the negative correlation between the oil
markets and the natural gas market in Japan, while all the other cor­
4.2. Dynamic spillover results
relations among the energy markets are positive.
Fig. 1 also presents the time series behavior of the different series.
In this section, we apply the methodology described above to
The most important noticeable observation is the rise in the uncertainty
investigate the dynamic spillovers across the different uncertainty
indices for the past six months in our series. The highest fluctuations are
indices and energy markets. To this end, we base our analysis on a TVP-
observed in the case of the EPU index compared to other indices. This
VAR. The different spillover indices are based on a 10-step-ahead fore­
variability coincides with the COVID-19 pandemic crisis, where the
cast.5 Tables 3 and 4 summarize the estimation results of the average
major economies experienced large fluctuations in their macro vari­
static and dynamic connectedness measures for each considered energy
ables, such as GDP, unemployment, and stock prices.
markets and uncertainty indices. We observe that own-index or energy
As shown in Fig. 1, the market uncertainties indices have consider­
market returns spillovers explain the highest share of forecast error
ably large swings observed from time to time. Focusing on the EPU, we
observe significant spikes during the 2001–2003 period when the
September 11 terrorist attack, the dot-com crisis and, the second Gulf 5
We run the TV-VAR model with lag one, and for robustness, we repeat the
War took place. Then, the index peaked again in 2007–2008 with the analysis based on many VAR lag orders and forecasting horizons, the results are
Global Financial Crisis (GFC) and the large cuts in interest rates for the similar.

6
A. Assaf et al. Resources Policy 72 (2021) 102112

variance, as the diagonal elements show higher values compared to the

off-diagonal elements. Also, Table 3 shows the total connectedness index
(TCI) measure. It represents the average influence that all variables have

Nat Gas
(Index)
on one variable’s forecast error variance throughout time. The TCI in all

1.00
markets is 52.88% in the case of static case, while 54.85% in the case of
dynamic connectedness scenario. This indicates that the energy markets
and market uncertainty indices are not independent of each other; that is
the average influence of market uncertainty on energy markets is
Nat Gas
(Japan)

approximately 53%. This large value indicates that the transmission of

0.233
1.00
market uncertainty represented by EPU, WTU, EMU, and GPR market
spillovers is an important source of energy market fluctuations.
Looking again at Table 4 and considering which market uncertainty
contributes the most to energy market fluctuation, we notice that the
EPU, followed by the WTU, contributes the most to the other markets.
(Europe)
Nat Gas

0.409
0.440
For example, EPU and WTU contribute 57.03% and 30.12% each to
1.00 other markets. Considering which energy sector contributes the most to
other markets, we notice that the oil market, on average, contributes
between 83% and 90% to the forecast error variance of all other mar­
Nat Gas

kets. This result confirms the dominance of oil in driving the uncertainty
0.186

0.093
0.942
(US)

1.00

indicators and the other energy prices fluctuation. Such a finding is not
surprising in the sense that oil is the most important energy source. At
this point, we notice that Dubai and Brent oil prices contribute better
than the WTI. A plausible explanation of this result is that the WTI crude
Coal (South

oil price mainly reflects the general equilibrium between US domestic

Africa)

demand and supply. However, the Brent and Dubai crude oil price de­
0.155
0.194

0.142
0.184
1.00

notes the benchmark price for Europe, Middle East, Russia, and North
Africa. In fact, the higher connectedness attributed to the Dubai and
Brent crude oil price may be due to the great degree of connection with
(Colombia)

the global economy as well as to the high degree of their globalization

(Yang, 2019). For the coal energy sector, the South African coal sector is
0.846
0.155
0.182

0.208
0.175
Coal

1.00

the most contributor by about 74%, followed by the Colombian coal

with about 68%. This result reflects the influential position of South
African’s coal price. In fact, South Africa is a significant participant in
global coal markets despite it is not the biggest producer. Eberhard
(Australia)

(2011) reports that South Africa’s coal industry is characterized by

0.547
0.675
0.065
0.075

0.004
0.075
Coal

1.00

low-cost producers compared to Colombian and Australian ones. This

characteristic makes South African coal price influential in the world
coal market. Moreover, South Africa is the largest coal export terminal,
− 0.071

and it is a potential swing producer.6 On the other hand, results show no

(WTI)

0.316
0.325
0.316
0.211
0.077

0.219
1.00
Oil

dominance of the Australian coal price in transmitting shocks to the

system despite its position as the largest exporter in the world. A plau­
sible explanation of this finding may be related to Australia’s strategy
(Dubai)

− 0.085

and most industrialized countries to reduce the use and production of

0.917
0.353
0.380
0.362
0.175
0.043

0.174
1.00
Oil

coal to mitigate pollution, especially with the world’s trend to reduce

the use of coal and replace it with clean energy.
Notes: The table presents the correlation matrix among the variables.

Considering how the shocks coming from others affect those markets,
− 0.091
(Brent)

1.000
0.967
0.936
0.371
0.359
0.346
0.173
0.055

0.175

we notice first that EMU receives the most contribution from other
Oil

markets by 52%, followed by the Economic Policy Uncertainty by 42%.

The oil market then receives about 70% contribution of shocks coming
− 0.178
− 0.168
− 0.181
− 0.232
− 0.250
− 0.269
− 0.091
− 0.131

− 0.047
− 0.104

from other markets, followed by the coal markets. The magnitude of

EMU

1.00

spillover contribution is qualitatively similar among the different series

considered for each market.
− 0.059
− 0.054
− 0.052
− 0.138

− 0.072
− 0.021
− 0.021

− 0.126
09.053

The bottom lines in Tables 3 and 4 show the total net spillovers for
0.004

0.062
WTU

1.00

each market. The results indicate that EMU and GPR, Coal Australia,
Natural Gas Europe, and Natural Gas Japan are net receivers of spillover
− 0.135
− 0.139
− 0.108
− 0.132
− 0.061
− 0.069

− 0.036

− 0.031
− 0.040
Correlation Matrix among the series.

from all others. In contrast, other markets and indices are net trans­
0.349
0.034

0.051
1.00
GPR

mitters for all others. Our results showing that the EPU and the WTU are
on average net transmitters of spillover to others is partially in line with
− 0.215
− 0.238
− 0.218
− 0.200
− 0.140
− 0.150
− 0.063
− 0.051

− 0.064

the findings of Wang and Lee (2020), who find that oil return acts as a
0.360
0.396
0.358

0.033
1.00
EPU

net recipient spillover from the EPU of oil-importing countries. Our

Coal (Colombia)

Nat Gas (Japan)

Coal (Australia)

Nat Gas (Index)

Coal (S Africa)
Nat Gas (US)
Oil (Dubai)

6
During the considered period of study, the coal export of the South African’s
(Europe)
Oil (Brent)

Oil (WTI)

Nat Gas

coals maintains a strategic position in the world since the first export coal
Table 2

WTU
EMU
GPR

destination of South Africa is India, one of the largest energy consumer in the
EPU

world.

7
A. Assaf et al. Resources Policy 72 (2021) 102112

Fig. 1. Monthly uncertainty indices evolution between January 2001 and July 2020.

findings at this point are also consistent with the views of Yin (2016) and interest rate used in the cash flow predictions. Thus, an increased in­
Yang (2019). They may be explained by the fact that the economic terest rate lead to equity markets instability and increased uncertainty in
policy uncertainty is transmitted to energy prices generally through the equity markets.
some channels, the most important is related to exchange rates policy To see the average of connectedness between the energy markets and
(Wang and Lee, 2020).7 The oil supply and real economic shocks can market uncertainties, we map the Total Connectedness Index (TCI)
also contribute significantly to the explanation of the linkage between through time. Fig. 2 presents the results for the dynamic total connect­
the EPU and oil returns (Yin, 2016). edness index (TCI). We observe large variations on the total connect­
Regarding the geopolitical risk uncertainty, results show negative edness index during our sample period. The total connectedness index is
net connectedness, indicating that this type of uncertainty is a spillover relatively high during the entire period. It exceeds the 60% level at
receiver from energy markets and other uncertainty measures. Such a observation 100, which coincides with the Global Financial Crisis of
result implies that the GPRs are influenced by energy markets, con­ 2008. Then TCI reached its lowest level with approximately 51% at
firming previous studies’ findings indicating a significant linkage be­ around observation 230, coinciding with the COVID-19 pandemic
tween energy prices and the GPRs, including Antonakakis et al. (2017a, outbreak in the first quarter of the year 2020.
b) and Qin et al. (2020), among others. The existing relationship be­ Our results are not surprising since the energy markets and market
tween the GPRs and energy prices may be related to the fact that uncertainties are interlinked and interdependent (Yin, 2016; Antona­
geopolitical instability, in its important part, is historically associated kakis et al., 2017a; Yang, 2019; Wang and Lee, 2020; Qin et al., 2020)
with the oil-producing regions such as the Middle East (Gulf Wars, and that crisis period may impact their interdependence. The outbreak
Saudi/Iran conflict) and Africa (Libyan conflicts, …). Therefore, the of the coronavirus pandemic and its global impact on the world econ­
geopolitical risk is discounted by the energy market participants in these omies and commodity markets may increase economic, geopolitical,
regions in the sense that energy market players act similarly in the oil or trade, and equity uncertainties and then induce negative changes in
other energy markets and synchronize their trading activity following a investor’s sentiments. That will also affect investment decisions and
geopolitical risk shock occurrence (Antonakakis et al., 2017a). energy price dynamics over time. Such unexpected risk shocks will affect
The equity uncertainty index EMU presents the highest negative investors’ appetite towards risks and their investment prospects (Sharif
value of average net spillover connectedness by a value of − 37.4%. This et al., 2020; Salisu and Adediran, 2020).
result implies that the uncertainty regarding equity markets receives This dynamical behavior of the TCI provides evidence that during
socks intensively from the energy markets and other categorical uncer­ our sample period, some financial and economic events may have a
tainty. This finding can be justified by the significant linkage between differential impact on these markets and their interconnectedness. As a
energy and equity markets, related mainly to investors’ sentiment fac­ result, we consider the time-varying behavior of connectedness mea­
tors. In fact, when a shock in energy (especially for oil) markets occurs, sures when analyzing the transmission of spillovers mechanism between
investors respond by interpreting these socks as bad news. As a result, energy markets and market uncertainties to have a better understating
investors sell their stocks, leading to stock market depreciation. Conse­ of these spillovers, especially during important critical periods.
quently, equity markets become riskier and present high uncertainty.
Another factor that can explain our results is related to the interest rate. 4.3. Robustness analysis
In fact, energy markets’ shocks lead to inflation and instability of the
To verify our results’ robustness and provide more valuable infor­
mation on the dynamic connectedness between energy prices and the
7 different uncertainty measures, further analysis is conducted. We
By giving a more detailed analysis of the connectedness between oil returns
investigate the dynamic connectedness between each uncertainty mea­
and the EPU, Wang and Lee (2020) conducted a disaggregated EPU- based
analysis and show that the EPU spillover is mainly related to fiscal policy and sure and energy prices. This analysis is made in order to study the re­
exchange rate policy. Such a results can then highlight that the role of EPU as a sponses of energy markets as represented by the oil, coal, and gas market
transmitter is mainly due to fiscal and exchange rates related uncertainty, to uncertainty measures. Using the different energy prices and uncer­
which confirms the literature postulating that oil price is significantly linked to tainty measures individually, we can estimate the effects of each un­
exchange rate (Youssef and Mokni, 2020). certainty on the three energy markets. The Appendix includes the results

8
 Table 3
Static connectedness matrix.
A. Assaf et al.

EPU GPR WTU EMU Oil Oil Oil Coal Coal Coal (South Nat Gas Nat Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) Africa) (US) (Europe) (Japan) (Index)

EPU 61.92 8.031 12.51 0.315 3.856 4.386 3.977 2.268 0.124 0.317 0.136 0.443 1.557 0.1520.627 38.07
GPR 7.731 76.79 10.83 0.162 0.813 0.684 0.628 0.512 0.379 0.126 0.260 0.386 0.065 0.265 23.20
WTU 7.065 7.34 73.38 0.248 1.089 0.713 0.814 4.955 0.294 0.102 0.070 3.033 0.623 1.123 26.61
EMU 23.73 0.512 0.909 38.62 6.569 7.965 6.086 2.699 3.070 2.130 1.456 0.222 4.887 1.479 61.37
Oil (Brent) 1.349 1.167 0.555 0.362 29.25 27.84 25.57 3.276 3.479 3.191 1.338 0.455 0.776 1.517 70.74
Oil (Dubai) 1.648 1.375 0.467 0.239 27.55 29.710 24.50 2.896 3.911 3.444 1.335 0.433 0.968 1.996 70.29
Oil (WTI) 1.418 1.021 0.518 0.237 27.00 26.115 30.64 2.384 2.911 2.555 1.660 0.722 0.805 1.518 69.35
Coal (Australia) 1.508 1.261 2.508 0.658 6.027 6.345 4.619 40.159 2.925 19.51 1.873 0.219 0.587 1.486 59.84
Coal (Colombia) 1.225 0.372 0.394 0.833 5.831 7.084 4.981 10.064 13.46 28.22 1.664 0.152 0.564 1.754 62.87
Coal (S Africa) 1.530 0.465 0.484 0.764 5.198 6.237 4.541 14.178 25.49 37.08 1.933 0.236 0.099 41.23 54.06
Nat Gas (US) 0.223 0.080 0.079 0.149 2.033 2.193 2.665 0.242 1.758 1.376 45.93 1.762 0.273 13.91 44.53
Nat Gas (Europe) 1.317 0.084 1.491 0.658 1.315 1.366 1.722 1.134 4.431 3.984 5.641 55.48 7.486 5.361 38.15
Nat Gas (Japan) 3.422 0.221 0.531 0.676 1.806 1.743 1.763 1.648 4.230 3.810 2.762 10.181 61.84 41.66 58.33
Nat Gas (Index) 0.241 0.102 0.152 0.193 2.192 2.350 3.010 0.204 1.621 1.346 37.31 7.985 1.639 72.41 740.37
Contribution to 52.411 22.04 31.43 5.493 91.28 94.919 84.78 46.46 65.18 69.84 57.44 26.23 20.33 114.08 TCI
OTHERS
Contribution 114.33 98.83 104.8 44.119 120.54 2124.72 115.53 86.62 102.31 69.84 103.37 81.70 82.17 14.08 52.88
Including Own
Net Spillovers 14.335 − 1.165 4.812 − 55.88 20.54 24.629 15.53 − 13.38 2.318 106.92 3.373 − 18.29 − 17.82

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the uncertainty indices (EPU, GPR, WTU, and EMU) and energy markets returns. The variance decompositions are
obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

9
Table 4
Dynamic connectedness matrix.
EPU GPR WTU EMU Oil Oil Oil Coal Coal Coal (South Nat Gas Nat Gas Nat Gas Natl Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) Africa) (US) (Europe) (Japan) (Index)

EPU 58.48 9.077 10.57 3.609 3.180 3.922 3.516 2.720 1.108 1.712 0.750 1.029 0.590 0.732 42.51
GPR 11.95 63.24 9.862 2.695 1.558 1.479 1.685 1.008 0.808 0.862 1.425 1.108 0.462 1.950 36.75
WTU 9.73 6.113 75.47 0.429 0.753 0.799 0.822 1.340 1.104 0.7515 0.391 1.369 0.519 0.438 24.52
EMU 18.21 1.887 1.685 46.69 4.311 4.763 3.815 3.950 3.841 4.535 1.743 0.709 2.528 1.320 52.30
Oil (Brent) 2.046 1.284 0.424 0.838 28.40 26.19 24.25 3.764 3.649 3.466 2.251 0.566 0.856 1.995 71.59
Oil (Dubai) 2.750 1.526 0.663 0.680 25.89 28.31 23.00 3.741 3.885 3.762 2.060 0.539 1.304 1.872 71.68
Oil (WTI) 2.341 0.883 0.333 0.643 25.48 24.46 29.42 3.051 3.074 2.873 2.775 0.892 1.020 2.733 70.57
Coal (Australia) 1.902 0.885 0.781 1.327 5.939 5.340 4.791 37.92 16.00 20.54 1.486 0.525 0.317 1.234 62.07
Coal (Colombia) 1.106 0.490 0.397 1.411 5.726 6.588 4.825 13.32 34.78 26.72 1.619 0.763 0.826 1.432 65.21
Coal (S Africa) 1.197 0.535 0.555 1.546 5.120 5.925 4.357 17.60 24.42 34.34 1.682 0.759 0.417 1.538 65.66
Nat Gas (US) 0.795 0.649 0.299 0.410 2.283 2.766 3.549 0.954 1.827 1.774 43.29 1.403 0.693 39.30 56.70
Nat Gas (Europe) 1.804 0.413 2.666 0.856 1.316 1.339 1.691 1.114 3.176 2.802 4.327 55.87 11.3 11.23 44.12
Nat Gas (Japan) 2.614 0.435 1.573 1.034 3.100 3.392 3.199 1.644 3.927 3.223 2.261 12.27 56.34 4.96 43.65
Nat Gas (Index) 0.571 0.517 0.308 0.412 2.179 2.592 3.622 0.721 1.487 1.552 36.54 6.498 2.491 40.50 59.49
Contribution to 57.03 24.69 30.12 15.89 86.85 90.55 83.12 54.93 68.33 74.54 59.32 28.44 23.41 70.64 767.90
OTHERS
Contribution 114.51 87.93 105.59 62.59 115.26 118.86 112.55 92.85 103.11 108.88 102.61 84.31 79.75 111.14 TCI
Including Own
Net Spillovers 14.51 − 12.06 5.595 − 37.40 15.26 18.86 12.56 − 7.142 3.115 8.881 2.615 − 15.68 − 20.24 11.14 54.85

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the uncertainty indices (EPU, GPR, WTU, and EMU) and energy markets returns. The variance decompositions are
obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.
Resources Policy 72 (2021) 102112
A. Assaf et al. Resources Policy 72 (2021) 102112

Fig. 2. Monthly energy returns evolution between January 2001 and July 2020.

markets, which confirm the primordial role of energy markets as a po­

tential determinant of economic and trade policy uncertainty. This result
may due principally to behavioral factors regarding the investors’
sentiment following shocks occurring in energy markets.
Turning to the equity and geopolitical risks policy uncertainty and its
individual connectedness analysis with energy markets. Results confirm
the robustness of the results found previously in Table 4. In fact, the
results of the dynamic connectedness in Tables A2 and A6 show that
both equity and geopolitical risk uncertainty again receive spillovers
that transmit, leading to the fact that these two types of uncertainty are
net spillovers receivers even the analysis is made individually. These
findings, robust to previous analysis in this study, confirm the role of the
EMU and GPRs as net receivers of the different shocks coming from the
energy markets.
Fig. 3. Total Connectedness Index based on the TV-VAR method. Overall, this study confirms that the different types of uncertainty are
receivers more than a transmitter of spillovers related to energy markets,
of the TVP-VAR model’s analysis of the effects of each uncertainty
indicating that energy prices are drivers of uncertainty in each category.
measure on the energy markets. Tables A1 to A8 include the results.
Starting with the EPU and the WTU indices, the results show that the
4.4. Dynamic connectedness and investors’ sentiment
economic and the trade policy uncertainty move from a net transmitter
to one receiver of spillovers compared to the results of Table 4. This is to
After running the TVP-VAR model and obtaining the different spill­
say, the EPU and WTU transmit information to the other uncertainty
over indices, we analyze the impact of the consumer sentiment index
indices more than to energy markets but receives spillovers from energy
(CSI) on the total connectedness and net spillover of each uncertainty
markets more than the other uncertainty types. Moreover, we find that
index and energy markets. The CSI is one of the most used indices as a
the EPU and WTU play a role of net spillovers receiver from energy
proxy of sentiment. It is based on the surveys of individuals, while the

10
A. Assaf et al. Resources Policy 72 (2021) 102112

Fig. 4. TO dynamic connectedness index.

former is extracted from several financial variables.

CIt = α0 + α1 CSIt + et (8)
The use of this index in our study is motivated by the large use of this
indicator in the literature. In fact, numerous studies consider the CSI as a
where CIt denotes the total connectedness or the net connectedness
leading indicator of real economic conditions (Acemoglu and Scott,
index between uncertainty and energy markets. CSIt represents the
1994; Howrey, 2001; Guo et al., 2020). It is employed to make policy
consumer sentiment index and et is the error term of the regression. A
decisions in many countries by offering qualitative information for
statistically significant parameter α1 indicates that the CSI exerts a sig­
presenting current conditions and predicting a household’s future eco­
nificant effect on the dynamic spillover index and then pushes the
nomic activity (Song and Shin, 2019). Moreover, Yao et al. (2012)
spillover between the considered variables.
indicate the importance of the CSI, which is considered as a monitored
To realize a robustness analysis and provide a more comprehensive
economic indicator, since it represents a robust measure of the consumer
picture on the effect of the investors’ sentiment on the dynamic
expectations and the changes in this index should contain important
connectedness between energy prices and uncertainty measures, we also
signals about future consumer behavior, as well as changes in the real
extend Eq. (7) to a quantile regression (QR) approach to analyze
economy and financial markets. In addition, some previous studies, such
whether the CSI exert an effect depending on the distribution level of the
as Carroll et al. (1994) and Howrey (2001), among others, show that
connectedness. This analysis is motivated by the fact that the different
consumer sentiment has incremental predictive power on the economic
connectedness measures move with varying levels. A QR analysis can
variables, including consumption and economic cycles.
clarify the type and strength of the effect of investors’ sentiment on these
Based on the regression model, we can test the effect of the investors’
connectedness measures. Furthermore, QR models have the advantage
sentiment on the dynamic (total and net) connectedness between un­
of fitting non-normal data series.8 Based on this approach, we estimate
certainty and energy markets. Such an analysis can finally extract
further information about the investor sentiment in driving the spillover
effects between the uncertainty and energy markets. Therefore, the ef­
8
fect of the investor sentiment on the different connectedness can be The Jarque-Bera test of normality is implemented on the connectedness
tested via the following regression model: measures as well as the CSI. We find that the normality is rejected for all series.
Due to space reasons, the results of this test are upon request.

11
A. Assaf et al. Resources Policy 72 (2021) 102112

Fig. 5. From dynamic connectedness index.

the following regression: therefore result in an overall negative impact. However, the positive
impact of consumer sentiment on energy prices can be explained by the
τ τ τ τ
Qt (CI) = α0 + α1 CSIt + et (9)
fact that energy prices are indicators of the phase in which the economy
goes through. The role of consumer sentiment might be vital in the
where Qτt (CI) represents the τth quantile order of the connectedness
dispersion of financial crises and economic uncertainty. Our results
measure. ατ1 measures the impact of the CSI on the τth distributional level confirm those of Nartea et al. (2020), who find that EPU premium is
of the connectedness measure. weaker following periods of high investor sentiment. This result is also
The estimation results of the parameter α1 and ατ1 are provided in in line with other previous studies, including Zhang (2018) and Rehman
Table 5. The estimation results emphasize a significant negative effect of and Apergis (2019), who find a significant relationship between in­
the sentiment on the total spillover among the considered uncertainty vestors’ sentiment and EPU.
indices and energy returns, indicating that a rise of sentiment level leads Regarding the positive effect of the investor’s sentiment on the net
to a decrease in the level of connectedness between these variables. spillover effects of the different energy markets, these results can be
However, the results for the net spillovers change remarkably according explained by the fact that an augmented investors’ sentiment is an
to uncertainty indices and energy markets. In fact, the CSI has a sig­ indication of a good economic situation since it has a predictive power of
nificant negative effect on the net connectedness of different uncertainty the real GDP and the personal consumption expenditure (Howrey,
measures, while the effect becomes significantly positive for the net 2001). Therefore, this situation is manifested by an increase in global
connectedness of the different all energy returns, except the Europe production, leading to an increase in energy demand, pushing energy
natural gas. prices to increase. As a result, energy markets gain the ability of shock
The negative impact of the consumer sentiment index on the total transmission, which resides mainly on the energy demand shocks
connectedness between energy markets and uncertainty indices means (Kilian, 2009; Kilian and Park, 2009). As examples of previous studies
that an augmented sentiment is an indication of good news, leading to confirming our findings, Qadan and Nama (2018) show that investor
the uncertainty decrease. As a result, the investors’ decisions become sentiment has a significant effect on oil prices and can explain part of its
more secure and are not influenced by the uncertainty. Therefore, this volatility.
latter loses its capacity to transmit shocks to energy markets. Spillover The QR model estimation results clearly confirm our findings for the
from a given index will be seen as bad news for other indices and OLS regression estimates. In fact, the parameter estimates have a similar

12
A. Assaf et al. Resources Policy 72 (2021) 102112

Fig. 6. Net dynamic connectedness index.

Table 5
Estimation of the effect of consumer sentiment index (CSI) on the total and net connectedness.
OLS regression Quantile regression

τ = 0.1 τ = 0.2 τ = 0.5 τ = 0.6 τ = 0.8 τ = 0.9

Connectedness measure coefficient Coefficient

Total connectedness − 0.1297*** − 0.067*** − 0.062*** − 0.146*** − 0.162*** − 0.203*** − 0.199***

Net uncertainty
Net EMU − 0.0202*** − 0.010*** − 0.017*** − 0.011*** − 0.041*** − 0.024*** − 0.022***
Net EPU − 0.0548*** − 0.051*** − 0.057*** − 0.056*** − 0.057*** − 0.061*** − 0.058***
Net WTI − 0.0213*** − 0.035*** − 0.041*** − 0.021*** − 0.018*** − 0.010*** − 0.010***
Net GPR − 0.0087*** − 0.013*** − 0.002 − 0.014 − 0.011 − 0.006 − 0.005**
Net energy
Net Oil (WTI) 0.0030*** 0.004* 0.009 0.003* 0.002 0.004*** 0.006***
Net Oil (Dubai) 0.0092*** 0.004** 0.003 0.012 0.009 0.009**** 0.008***
Net Oil (Brent) 0.0162*** 0.006** 0.004* 0.022*** 0.023*** 0.022*** 0.021***
Net Coal (Australia) 0.0041* − 0.000 0.009*** 0.010** 0.01** 0.0015 0.004
Net Coal (Colombia) 0.0216*** 0.007 0.018*** 0.021*** 0.023*** 0.024*** 0.024***
Net Coal (South Africa) 0.0185*** 0.021*** 0.018*** 0.017*** 0.018*** 0.017*** 0.018***
Net Natural Gas (Japan) 0.0147*** 0.008*** 0.008*** 0.007*** 0.008*** 0.029*** 0.031***
Net Natural Gas (Europe) 0.0013 0.008*** 0.005** − 0.0004 − 0.003 − 0.004 − 0.001
Net Natural Gas (US) 0.0031** 0.005*** 0.007 − 0.0009 − 0.0026 0.0003 0.003***
Net Natural Gas (Index) 0.0133*** 0.013*** 0.018*** 0.012*** 0.011*** 0.010*** 0.008***

Notes: ***, **, * indicates significance at 1%, 5% and 10% respectively.

13
A. Assaf et al. Resources Policy 72 (2021) 102112

sign of the impact of investors’ sentiment on the dynamic connected­ then dropping to a level of 51%, coinciding with the COVID-19
ness. On the other hand, other than its robustness analysis gain, the QR pandemic outbreak. However, the net spillover indices generally find
model provides a more detailed picture of the impact of CSI on CI by that the EPU and the WTU transmit more than receiving the information
showing that generally, the impact of the investors sentiment on the on average. However, the EMU and GPR are net receivers of spillover
dynamic connectedness varies across distributional levels, indicating effects.
that more attention should be accorded to this relationship during the Our findings have important implications for policymakers and in­
high level of the total connectedness between energy markets and un­ vestors at some points. In fact, we globally find that uncertainty related
certainty in which the effect of the CSI is more pronounced. to the economic, trade, equity markets, and geopolitical risks play a
crucial role in energy markets and can destabilize energy prices strongly.
5. Conclusion This finding implies for policymakers that attention should be accorded
to the responses of energy prices to policy uncertainty shocks, which can
While the nexus of the uncertainty-energy markets has been well be crystallized in the form of a comb rise or decrease in energy prices.
examined in the literature, almost previous studies focused on one or at Therefore, policymakers’ intervention regarding decision-making in
least two categories of uncertainty measures and supposed only the oil energy markets should consider the uncertainty levels regarding eco­
prices to represent energy markets. The connectedness regarding other nomic, trade, equity markets, and geopolitical risks. The different types
uncertainty types still awaits identification and analysis. This paper of uncertainty should be implemented to stabilize energy prices, and
investigated the dynamic connectedness between the most important more attention should be paid to the economic policy uncertainty as to
uncertainty measures (EPU, WTU, GPR, and EMU) and most energy the best category that influences the dynamics of energy markets.
prices (oil prices, coal prices, and natural gas prices) to address this Regarding investors, given that our results clearly show that uncer­
limitation. We use the recent approach combining the time-varying tainty is a key factor that influences energy prices evolution, they should
parameters VAR (TVP-VAR) models and the connectedness measures account for uncertainty when planning investment decisions regarding
of Diebold and Yilmaz (2009, 2012, 2014). portfolio design and risk management analysis. In this context, with the
Based on monthly data over the period spanning from January 2001 financialization implying the use of energy prices in the financial in­
and July 2020, the results show that the different uncertainty shocks vestment, it is recommended to investors to pay more attention to the
play a significant role in energy returns. The influence of the investi­ EPU than other uncertainty types of uncertainty and to oil markers in
gated uncertainty on energy markets is approximately important, time- relation to uncertainty.
varying, and indicates an interactive connection. The EPU is the most Finally, we find that the investors’ sentiment has a significant impact
contributor to energy markets, followed by the World Trade Uncer­ on the connectedness between energy prices and uncertainty also have
tainty. By focusing on the energy markets, results show that the oil important policy implications. In fact, an increase in the investors’
market is the most contributor to the uncertainty, followed by the coal sentiment leads to a decrease in the total spillovers. This finding implies
market. that when investors are optimist, it can indicate the ability of energy
Regarding the received spillovers from other variables, the Equity prices as hedging and safe-haven assets against the uncertainty. Besides,
Market uncertainty appears as the most receiver of spillover effects from low connectedness can offer investors an opportunity for beneficial
the energy markets, followed by the EPU by 42%. Moreover, the results diversification by mixing energy prices in their portfolios during high
highlight that the total connectedness index is relatively high, exceeding uncertainty periods.
60% at a point that coincides with the Global Financial Crisis of 2008,

Table A1
Static Connectedness Matrix for EMU

EMU Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

EMU 47.799 9.547 10.763 9.019 6.796 3.716 5.652 1.689 0.419 2.998 1.603 52.201
Oil (Brent) 0.925 28.342 27.519 25.768 8.413 2.18 3.68 1.202 0.238 0.452 1.279 71.658
Oil (Dubai) 0.64 27.407 29.376 25.274 7.605 2.555 3.902 1.201 0.182 0.573 1.284 70.624
Oil (WTI) 0.711 26.91 26.591 29.619 7.038 1.937 3.22 1.469 0.368 0.444 1.694 70.381
Coal 2.115 13.032 12.852 10.995 37.839 6.201 14.009 1.452 0.047 0.282 1.178 62.161
(Australia)
Coal 1.339 5.965 7.203 5.461 7.641 38.928 29.1 1.825 0.316 0.475 1.747 61.072
(Colombia)
Coal 1.54 6.585 7.695 6.136 12.07 24.218 37.124 2.042 0.505 0.04 2.045 62.876
(S. Africa)
Nat Gas 0.2 1.878 2.019 2.433 0.386 1.697 1.469 46.617 1.581 0.248 41.472 53.383
(US)
Nat. Gas 1.128 4.433 4.251 4.452 5.049 3.325 3.979 5.508 51.058 3.185 13.632 48.942
(Europe)
Nat Gas (Japan) 1.792 2.157 2.199 2.072 1.889 4.175 4.311 2.517 7.515 66.14 5.231 33.86
Nat Gas (index) 0.338 2.179 2.274 2.878 0.544 1.641 1.439 37.083 8.316 1.405 41.904 58.096
Contribution TO others 10.728 100.093 103.366 94.489 57.43 51.646 70.761 55.988 19.486 10.102 71.165 645.25
Contribution including 58.527 128.435 132.742 124.108 95.269 90.573 107.885 102.605 70.544 76.242 113.069 TCI
own
Net spillovers − 41.47 28.435 32.742 24.108 − 4.731 − 9.427 7.885 2.605 − 29.456 − 23.758 13.069 58.659

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the Equity Market Index (EMI) index and energy markets returns
from static model. The variance decompositions are obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

14
A. Assaf et al. Resources Policy 72 (2021) 102112

Table A2
Dynamic Connectedness Matrix-EMU

EMU Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

EMU 60.065 5.293 5.592 4.821 5.279 5.37 6.327 1.626 1.332 2.838 1.457 39.935
Oil (Brent) 0.89 28.738 26.789 25.144 6.472 3.513 3.849 1.717 0.516 0.804 1.566 71.262
Oil (Dubai) 0.84 26.831 28.908 24.266 6.393 3.73 4.186 1.642 0.534 1.104 1.565 71.092
Oil (WTI) 0.87 26.234 25.267 29.422 5.578 3.065 3.387 2.184 0.868 0.863 2.261 70.578
Coal 1.537 9.302 9.446 7.874 37.546 12.546 18.427 1.307 0.516 0.342 1.157 62.454
(Australia)
Coal 2.014 6.247 7.062 5.421 11.35 36.003 27.277 1.522 0.926 0.725 1.453 63.997
(Colombia)
Coal 2.004 6.107 6.943 5.443 16.286 24.082 34.811 1.576 0.861 0.357 1.531 65.189
(S. Africa)
Nat Gas 0.512 2.035 2.313 2.97 0.811 1.744 1.715 45.097 1.537 0.717 40.549 54.903
(US)
Nat. Gas 1.793 1.394 1.461 1.793 1.137 3.781 3.256 4.229 58.197 10.626 12.334 41.803
(Europe)
Nat Gas (Japan) 1.915 3.52 3.696 3.538 2.102 4.219 3.61 2.092 11.583 58.67 5.054 41.33
Nat Gas (index) 0.587 2.02 2.27 3.128 0.667 1.539 1.6 36.955 7.34 2.552 41.341 58.659
Contribution TO others 12.962 88.984 90.839 84.399 56.073 63.589 73.634 54.85 26.015 20.929 68.929 641.2
Contribution including 73.027 117.723 119.747 113.82 93.619 99.592 108.445 99.947 84.212 79.599 110.27 TCI
own
Net spillovers − 26.97 17.723 19.747 13.82 − 6.381 − 0.408 8.445 − 0.053 − 15.788 − 20.401 10.27 58.291

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the Equity Market Index (EMI) and energy markets returns from
dynamic model. The variance decompositions are obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

Table A3
Static Connectedness Matrix-EPU

EPU Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

EPU 55.644 9.831 10.703 9.711 6.763 0.372 1.188 0.504 1.934 2.425 0.926 44.356
Oil (Brent) 1.631 28.097 27.264 25.529 8.52 2.125 3.77 1.094 0.261 0.5 1.212 71.903
Oil (Dubai) 2.108 26.949 28.854 24.845 7.632 2.503 3.983 1.101 0.206 0.598 1.221 71.146
Oil (WTI) 1.681 26.638 26.314 29.271 7.112 1.868 3.262 1.361 0.382 0.496 1.615 70.729
Coal 0.878 13.398 13.133 11.227 38.097 6.078 14.379 1.316 0.044 0.337 1.113 61.903
(Australia)
Coal 1.618 6.338 7.615 5.789 7.859 38.377 28.351 1.742 0.301 0.367 1.644 61.623
(Colombia)
Coal 1.538 7.117 8.26 6.565 12.545 23.343 36.433 1.887 0.389 0.059 1.865 63.567
(S. Africa)
Nat Gas 0.079 1.732 1.882 2.306 0.298 1.702 1.399 46.948 1.624 0.266 41.763 53.052
(US)
Nat. Gas 0.861 4.626 4.452 4.64 5.171 3.058 3.717 5.45 51.441 2.984 13.6 48.559
(Europe)
Nat Gas (Japan) 6.067 2.463 2.484 2.394 2.203 3.315 3.468 2.275 7.147 63.349 4.833 36.651
Nat Gas (index) 0.144 2.116 2.219 2.826 0.511 1.599 1.36 37.307 8.365 1.393 42.161 57.839
Contribution TO others 16.605 101.207 104.326 95.833 58.613 45.962 64.877 54.036 20.653 9.423 69.792 641.33
Contribution including 72.249 129.304 133.18 125.104 96.709 84.339 101.311 100.984 72.094 72.773 111.953 TCI
own
Net spillovers − 27.75 29.304 33.18 25.104 − 3.291 − 15.661 1.311 0.984 − 27.906 − 27.227 11.953 58.303

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the Economic Policy Uncertainty (EPU) and energy markets returns
from static model. The variance decompositions are obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

15
A. Assaf et al. Resources Policy 72 (2021) 102112

Table A4
Dynamic Connectedness Matrix-EPU

EPU Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

EPU 74.15 4.113 5.113 4.491 2.774 1.719 2.305 0.622 3.005 0.775 0.934 25.851
Oil (Brent) 1.938 28.452 26.649 24.756 6.435 3.256 3.801 1.653 0.531 1.038 1.491 71.548
Oil (Dubai) 2.527 26.453 28.447 23.729 6.286 3.475 4.046 1.612 0.524 1.384 1.517 71.553
Oil (WTI) 2.098 25.827 24.965 29.173 5.553 2.829 3.265 2.163 0.853 1.064 2.212 70.827
Coal 1.6 9.3 9.536 7.916 37.719 12.283 18.23 1.333 0.504 0.394 1.184 62.281
(Australia)
Coal 1.765 6.094 7.023 5.296 11.354 36.161 27.6 1.577 0.94 0.668 1.523 63.839
(Colombia)
Coal 1.593 6.154 7.016 5.429 16.221 24.208 34.985 1.624 0.831 0.352 1.588 65.015
(S. Africa)
Nat Gas 0.545 1.91 2.278 2.947 0.777 1.687 1.594 45.314 1.529 0.627 40.793 54.686
(US)
Nat. Gas 2.882 1.292 1.386 1.637 1.075 3.528 3.013 4.189 58.837 9.92 12.241 41.163
(Europe)
Nat Gas (Japan) 3.277 3.436 3.723 3.46 1.979 3.839 3.306 1.963 11.589 58.592 4.837 41.408
Nat Gas (index) 0.576 1.954 2.251 3.108 0.669 1.519 1.535 37.205 7.271 2.357 41.556 58.444
Contribution TO others 18.8 86.533 89.939 82.769 53.121 58.341 68.697 53.942 27.576 18.578 68.319 626.62
Contribution including 92.95 114.985 118.386 111.941 90.841 94.502 103.683 99.255 86.413 77.17 109.875 TCI
own
Net spillovers − 7.051 14.985 18.386 11.941 − 9.159 − 5.498 3.683 − 0.745 − 13.587 − 22.83 9.875 56.965

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the Economic Policy Uncertainty (EPU) and energy markets returns
from dynamic model. The variance decompositions are obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

Table A5
Static Connectedness Matrix-GPR

GPR Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

GPR 96.019 0.782 0.692 0.603 0.494 0.054 0.055 0.036 1.092 0.051 0.121 3.981
Oil (Brent) 1.094 28.346 27.442 25.73 8.468 1.96 3.556 1.194 0.247 0.681 1.283 71.654
Oil (Dubai) 1.306 27.229 29.149 25.099 7.577 2.339 3.778 1.216 0.174 0.827 1.306 70.851
Oil (WTI) 1.014 26.83 26.455 29.593 7.024 1.75 3.11 1.479 0.37 0.658 1.716 70.407
Coal 1.025 13.374 13.052 11.166 38.908 5.66 13.925 1.358 0.036 0.411 1.086 61.092
(Australia)
Coal 0.535 5.615 6.845 5.199 7.323 40.086 29.796 1.952 0.338 0.483 1.827 59.914
(Colombia)
Coal 0.423 6.548 7.665 6.116 12.116 24.5 37.896 2.11 0.488 0.047 2.09 62.104
(S. Africa)
Nat Gas 0.048 1.847 2.02 2.431 0.307 1.763 1.442 46.665 1.661 0.298 41.518 53.335
(US)
Nat. Gas 0.512 4.117 3.917 4.193 4.806 3.339 3.952 5.706 52.147 3.325 13.985 47.853
(Europe)
Nat Gas (Japan) 0.287 1.593 1.576 1.524 1.576 4.316 4.22 2.646 8.141 68.668 5.452 31.332
Nat Gas (index) 0.01 2.135 2.252 2.861 0.464 1.693 1.427 37.185 8.436 1.501 42.035 57.965
Contribution TO others 6.254 90.072 91.917 84.923 50.156 47.374 65.26 54.882 20.983 8.283 70.383 590.49
Contribution including 102.27 118.418 121.066 114.515 89.064 87.46 103.156 101.547 73.13 76.951 112.418 TCI
own
Net spillovers 2.273 18.418 21.066 14.515 − 10.936 − 12.54 3.156 1.547 − 26.87 − 23.049 12.418 53.681

16
A. Assaf et al. Resources Policy 72 (2021) 102112

Notes: This table reports the variance decompositions for the esti­
mated TVP-VAR model addressing the Geopolitical Index (GPR) and
energy markets returns from static model. The variance decompositions
are obtained based on 10-step-ahead forecasts and a lag of order 1 for
the TVP-VAR model.

Table A6
Dynamic Connectedness Matrix-GPR

GPR Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

GPR 84.9 2.121 2.011 2.1 0.846 0.646 0.581 2.045 1.501 0.953 2.296 15.1
Oil (Brent) 1.289 28.415 26.451 24.779 6.52 3.416 3.885 2.042 0.494 0.845 1.864 71.585
Oil (Dubai) 1.636 26.441 28.568 24.036 6.352 3.629 4.156 1.831 0.528 1.078 1.745 71.432
Oil (WTI) 1.075 26.034 25.245 29.349 5.603 2.934 3.332 2.306 0.837 0.912 2.374 70.651
Coal 1.063 9.393 9.53 7.9 37.865 12.413 18.317 1.41 0.518 0.339 1.251 62.135
(Australia)
Coal 0.573 6.058 6.942 5.166 11.482 36.805 28.032 1.629 0.998 0.739 1.577 63.195
(Colombia)
Coal 0.531 6.059 6.902 5.282 16.414 24.625 35.606 1.705 0.865 0.343 1.669 64.394
(S. Africa)
Nat Gas 0.542 2.261 2.494 3.086 0.815 1.658 1.61 44.944 1.483 0.629 40.479 55.056
(US)
Nat. Gas 0.516 1.406 1.464 1.816 1.098 3.772 3.208 4.25 59.671 10.492 12.307 40.329
(Europe)
Nat Gas (Japan) 0.626 3.603 3.788 3.66 2.119 4.272 3.628 2.092 11.682 59.486 5.044 40.514
Nat Gas (index) 0.401 2.293 2.488 3.288 0.702 1.535 1.601 36.929 7.108 2.422 41.234 58.766
Contribution TO others 8.253 85.669 87.314 81.112 51.949 58.9 68.35 56.24 26.015 18.751 70.604 613.16
Contribution including 93.15 114.084 115.883 110.461 89.814 95.705 103.957 101.183 85.686 78.237 111.838 TCI
own
Net spillovers − 6.848 14.084 15.883 10.461 − 10.186 − 4.295 3.957 1.183 − 14.314 − 21.763 11.838 55.742

Notes: This table reports the variance decompositions for the esti­
mated TVP-VAR model addressing the Geopolitical Index (GPR) and
energy markets returns from dynamic model. The variance de­
compositions are obtained based on 10-step-ahead forecasts and a lag of
order 1 for the TVP-VAR model.

Table A7
Static Connectedness Matrix-WTU

WTU Oil (Brent) Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

WTU 86.48 1.342 0.841 0.959 5.75 0.391 0.105 0.102 3.186 0.53 0.315 13.519
Oil (Brent) 0.562 30.212 28.756 26.383 3.67 3.466 3.248 1.213 0.444 0.699 1.346 69.788
Oil (Dubai) 0.478 28.678 30.916 25.413 3.309 3.872 3.484 1.195 0.418 0.878 1.36 69.084
Oil (WTI) 0.518 27.835 26.983 31.397 2.722 2.975 2.668 1.572 0.72 0.712 1.898 68.603
Coal 2.476 6.728 7.099 5.162 40.878 13.798 19.709 1.844 0.24 0.522 1.545 59.122
(Australia)
Coal 0.337 5.864 7.063 5.078 10.396 37.961 29.113 1.727 0.187 0.707 1.567 62.039
(Colombia)
Coal 0.434 5.334 6.34 4.7 14.58 26.24 38.098 1.969 0.279 0.186 1.839 61.902
(S. Africa)
Nat Gas 0.058 1.814 1.93 2.514 0.211 1.792 1.35 46.435 1.891 0.309 41.696 53.565
(US)
Nat. Gas 1.242 0.97 0.965 1.454 1.094 4.902 4.536 5.972 56.196 8.342 14.327 43.804
(Europe)
Nat Gas (Japan) 0.727 1.517 1.339 1.503 1.693 4.907 4.714 2.949 11.013 63.915 5.723 36.085
Nat Gas (index) 0.121 1.979 2.091 2.857 0.18 1.695 1.383 37.641 8.217 1.816 42.017 57.983
Contribution TO others 6.954 82.06 83.408 76.023 43.605 64.037 70.31 56.184 26.595 14.702 71.616 595.5
Contribution including 93.44 112.272 114.324 107.42 84.483 101.998 108.408 102.619 82.791 78.617 113.634 TCI
own
Net spillovers − 6.565 12.272 14.324 7.42 − 15.517 1.998 8.408 2.619 − 17.209 − 21.383 13.634 54.136

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the World Trade Uncertainty (WTI) and energy markets returns
from statistic model. The variance decompositions are obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

17
A. Assaf et al. Resources Policy 72 (2021) 102112

Table A8
Dynamic Connectedness Matrix-WTU

WTU Oil Oil Oil Coal Coal Coal Nat Gas Nat. Gas Nat Gas Nat Gas FROM
(Brent) (Dubai) (WTI) (Australia) (Colombia) (S. (US) (Europe) (Japan) (Index)
Africa)

WTU 89.48 0.992 1 0.957 1.706 1.391 0.904 0.464 1.934 0.604 0.565 10.517
Oil (Brent) 0.362 29.56 27.373 25.389 4.361 3.981 3.658 1.994 0.579 0.945 1.798 70.44
Oil (Dubai) 0.522 27.436 29.81 24.429 4.347 4.217 3.905 1.809 0.573 1.257 1.694 70.19
Oil (WTI) 0.286 26.729 25.651 30.5 3.54 3.407 3.039 2.409 0.959 0.995 2.486 69.5
Coal 0.825 6.646 7.092 5.365 39.087 16.413 21.03 1.414 0.561 0.328 1.24 60.913
(Australia)
Coal 0.268 6.205 7.069 5.171 13.704 35.601 27.494 1.55 0.744 0.771 1.423 64.399
(Colombia)
Coal 0.527 5.418 6.215 4.549 18.19 25.314 35.459 1.607 0.76 0.438 1.523 64.541
(S. Africa)
Nat Gas 0.226 2.137 2.355 3.047 0.838 1.736 1.627 44.947 1.65 0.686 40.753 55.053
(US)
Nat. Gas 3.313 1.2 1.206 1.626 1.225 3.467 3.017 4.385 56.607 12.208 11.748 43.393
(Europe)
Nat Gas (Japan) 1.879 3.41 3.562 3.452 1.894 4.349 3.54 1.914 13.155 58.181 4.663 41.819
Nat Gas (index) 0.329 2.117 2.276 3.201 0.675 1.486 1.499 37.356 7.105 2.508 41.447 58.553
Contribution TO others 8.537 82.289 83.799 77.186 50.479 65.761 69.713 54.902 28.02 20.74 67.892 609.32
Contribution including 98.02 111.849 113.608 107.687 89.567 101.362 105.173 99.848 84.627 78.921 109.339 TCI
own
Net spillovers − 1.981 11.849 13.608 7.687 − 10.433 1.362 5.173 − 0.152 − 15.373 − 21.079 9.339 55.393

Notes: This table reports the variance decompositions for the estimated TVP-VAR model addressing the World Trade Uncertainty (WTI) and energy markets returns
from a dynamic model. The variance decompositions are obtained based on 10-step-ahead forecasts and a lag of order 1 for the TVP-VAR model.

Declaration of competing interest Asai, M., Gupta, R., McAleer, M., 2020. Forecasting volatility and co-volatility of crude
oil and gold futures: effects of leverage, jumps, spillovers, and geopolitical risks. Int.
J. Forecast. 36 (3), 933–948.
The authors declare that they have no known competing financial Aye, G.C., 2018. Causality between economic policy uncertainty and real housing returns
interests or personal relationships that could have appeared to influence in emerging economies: a cross-sample validation approach. Cogent Economics &
the work reported in this paper. Finance 6 (1), 1473708.
Aye, G.C., Clance, M.W., Gupta, R., 2019. The effect of economic uncertainty on the
housing market cycle. J. Real Estate Portfolio Manag. 25 (1), 67–75.
References Bahmani-Oskooee, M., Harvey, H., Niroomand, F., 2018. On the impact of policy
uncertainty on oil prices: an asymmetry analysis. Int. J. Financ. Stud. 6, 1–11.
Acemoglu, D., Scott, A., 1994. Consumer confidence and rational expectations: are Baker, S.R., Bloom, N., Davis, S.J., 2016. Measuring economic policy uncertainty. Q. J.
agents’ beliefs consistent with the theory. Econ. J. 104 (422), 1–19. Econ. 131 (4), 1593–1636.
Ahir, Hites, Bloom, Nicholas, Furceri, Davide, 2019. The World Uncertainty Index. Baker, S.B., Bloom, N., Davis, S.J., 2019a. The extraordinary rise in trade policy
Stanford mimeo. Retrieved from. [Link] uncertainty. [Link] 17 (September).
mimeo_10_29.pdf. Baker, S.B., Bloom, N., Davis, S.J., Kost, K., 2019b. Policy News and Equity Market
Ajmi, A.N., Gupta, R., Kanda, P.T., 2014. Causality between economic policy uncertainty Volatility. NBER Working Paper 25720.
across countries: evidence from linear and nonlinear tests. Front. Fam. Econ. 11, Balcilar, M., Bekiros, S., Gupta, R., 2017. The role of news-based uncertainty indices in
73–102. predicting oil markets: a hybrid nonparametric quantile causality method. Empir.
Ajmi, A.N., Aye, G.C., Balcilar, M., El Montasser, G., Gupta, R., 2015. Causality between Econ. 53, 879–889.
US economic policy and equity market uncertainties: evidence from linear and Bekiros, S., Gupta, R., Majumdar, A., 2016. Incorporating economic policy uncertainty in
nonlinear tests. J. Appl. Econ. 18 (2), 225–246. US equity premium models: a nonlinear predictability analysis. Finance Res. Lett. 18,
Aloui, R., Gupta, R., Miller, S.M., 2016. Uncertainty and crude oil returns. Energy Econ. 291–296.
55, 92–100. Berthou, A., Jardet, C., Siena, D., Szczerbowicz, U., 2019. The Macroeconomic
André, C., Bonga-Bonga, L., Gupta, R., Muteba Mwamba, J.W., 2017. Economic policy Implications of a Global Trade War. [Link], 8 February.
uncertainty, US real housing returns and their volatility: a nonparametric approach. Bianconi, M., Esposito, F., Sammon, M., 2019. Trade Policy Uncertainty and Stock
J. R. Estate Res. 39 (4), 493–514. Returns. Available at SSRN 3340700.
Anoruo, E., Akpom, U., Nwoye, Y., 2017. Dynamic relationship between economic policy Bloom, N., 2009. The impact of uncertainty shocks. Econometrica 77 (3), 623–685.
uncertainty and housing market returns in Japan. Journal of International Business Bouoiyour, J., Selmi, R., Hammoudeh, S., Wohar, M.E., 2019. What are the categories of
and Economics 5 (2), 28–37. geopolitical risks that could drive oil prices higher? Acts or threats? Energy Econ. 84,
Antonakakis, N., Chatziantoniou, I., Filis, G., 2013. Dynamic co-movements of stock 104523.
market returns, implied volatility and policy uncertainty. Econ. Lett. 120, 87–92. Brogaard, J., Detzel, A., 2015. The asset-pricing implications of government economic
Antonakakis, N., Chatziantoniou, I., Filis, G., 2014. Dynamic spillovers of oil price shocks policy uncertainty. Manag. Sci. 61 (1), 3–18.
and economic policy uncertainty. Energy Econ. 44, 433–447. Caldara, D., Iacoviello, M., 2018. Measuring geopolitical risk. In: International Finance
Antonakakis, N., Chatziantoniou, I., Filis, G., 2017b. Oil shocks and stock markets: Discussion Paper, vol. 1222, pp. 1–66.
dynamic connectedness under the prism of recent geopolitical and economic unrest. Caldara, D., Iacoviello, M., Molligo, P., Prestipino, A., Raffo, A., 2020. The economic
Int. Rev. Financ. Anal. 50 (C), 1–26. effects of trade policy uncertainty. J. Monetary Econ. 109, 38–59.
Antonakakis, N., Chatziantoniou, I., Gabauer, D., 2020. Refined measures of dynamic Carroll, Christopher D., Fuhrer, Jeffrey C., Wilcox, David W., 1994. Does consumer
connectedness based on time-varying parameter vector autoregressions. J. Risk sentiment forecast household spending? If so, why? Am. Econ. Rev. 84, 1397–1408.
Financ. Manag. 13 (4), 84. Chen, J., Jin, F., Ouyang, G., Ouyang, J., Wen, F., 2019. Oil price shocks, economic
Antonakakis, N., Cunado, J., Filis, G., Gabauer, D., Perez de Gracia, F., 2018. Oil policy uncertainty and industrial economic growth in China. PloS One 14 (5),
volatility, oil and gas firms and portfolio diversification. Energy Econ. 70, 499–515. e0215397.
Antonakakis, N., Gabauer, D., 2017. Refined Measures of Dynamic Connectedness Based Chen, X., Sun, X., Wang, J., 2019b. Dynamic spillover effect between oil prices and
on TVP-VAR. MPRA Paper No. 78282. economic policy uncertainty in BRIC countries: a wavelet-based approach. Emerg.
Antonakakis, N., Gupta, R., Kollias, C., Papadamou, S., 2017a. Geopolitical risks and the Mark. Finance Trade 1–15. [Link]
oil-stock nexus over 1899–2016. Finance Res. Lett. 165–173. Christou, C., Gabauer, D., Gupta, R., 2020. Time-Varying impact of uncertainty shocks on
Arouri, M., Estay, C., Rault, C., Roubaud, D., 2016. Economic policy uncertainty and macroeconomic variables of the United Kingdom: evidence from over 150 years of
stock markets: long-run evidence from the US. Finance Res. Lett. 18, 136–141. monthly data. Finance Res. Lett. 37, 101363.
Arouri, M., Rault, C., Teulon, F., 2014. Economic policy uncertainty, oil price shocks and Cunado, J., Gupta, R., Lau, C.K.M., Sheng, X., 2019. Time-varying impact of geopolitical
GCC stock markets. Econ. Bull. 34 (3), 1822–1834. risks on oil prices. Defense and Peace Economics. [Link]
10242694.2018.1563854.
Demirer, R., Gupta, R., Ji, Q., Tiwari, A.K., 2019. Geopolitical risks and the predictability
of regional oil returns and volatility. OPEC Energy Rev 43, 342–361.

18
A. Assaf et al. Resources Policy 72 (2021) 102112

Diebold, F.X., Yilmaz, K., 2009. Measuring financial asset return and volatility spillovers, Ma, R., Zhou, C., Cai, H., Deng, C., 2019. The forecasting power of EPU for crude oil
with application to global equity markets. Econ. J. 119 (534), 158–171. return volatility. Energy Rep. 5, 866–873.
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional Maslyuk-Escobedo, S., Rotaru, K., Dokumentov, A., 2017. News sentiment and jumps in
measurement of volatility spillovers. Int. J. Forecast. 28 (1), 57–66. energy spot and futures markets. Pac. Basin Finance J. 45, 186–210.
Diebold, F.X., Yılmaz, K., 2014. On the network topology of variance decompositions: Mei, D., Ma, F., Liao, Y., Wang, L., 2020. Geopolitical Risk Uncertainty and Oil Future
measuring the connectedness of financial firms. J. Econom. 182 (1), 2014. Volatility: Evidence from MIDAS Models. Energy Economics, vol. 86. Elsevier. C.
Eberhard, A., 2011. The future of South African coal: market, investment and policy Nartea, G.V., Bai, H., Wu, J., 2020. Investor sentiment and the economic policy
challenges. Program on energy and sustainable development 1–44. uncertainty premium. Pac. Basin Finance J. 101438.
Feng, Y., Xu, D., Failler, P., Li, T., 2020. Research on the time-varying impact of Olasehinde-Williams, g., 2020. Is US trade policy uncertainty powerful enough to predict
economic policy uncertainty on crude oil price fluctuation. Sustainability 12 (16), global output volatility? J. Int. Trade Econ. Dev. [Link]
6523. 09638199.2020.1806912.
Gabauer, D., Gupta, R., 2018. On the transmission mechanism of country-specific and Peng, C., Zhu, H., Guo, Y., Chen, X., 2018. Risk spillover of international crude oil to
international economic uncertainty spillovers: evidence from a TVP-VAR China’s firms: evidence from granger causality across quantile. Energy Econ. 72,
connectedness decomposition approach. Econ. Lett. 171, 63–71. 188–199.
Gong, X., Lin, B., 2017. Forecasting the good and bad uncertainties of crude oil prices Plakandaras, V., Gupta, R., Wong, W.-K., 2019. Point and density forecasts of oil returns:
using a HAR framework. Energy Econ. 67, 315–327. the role of geopolitical risks. Resour. Pol. 62, 580–587.
Gong, X., Lin, B., 2018. Structural breaks and volatility forecasting in the copper futures Primiceri, G.E., 2005. Time varying structural vector autoregressions and monetary
market. J. Futures Mark. 38 (3), 290–339. policy. Rev. Econ. Stud. 72 (3), 821–852.
Gong, X., Lin, B., 2019. Modeling stock market volatility using new HAR-type models. Qadan, M., Nama, H., 2018. Investor sentiment and the price of oil. Energy Econ. 69,
Phys. Stat. Mech. Appl. 516, 194–211. 42–58 eneco.2017.10.035.
Gozgor, G., Tiwari, A.K., Demir, E., Akron, S., 2019. The relationship between Bitcoin Qin, Y., Hong, K., Chen, J., Zhang, Z., 2020. Asymmetric effects of geopolitical risks on
returns and trade policy uncertainty. Finance Res. Lett. 29, 75–82. energy returns and volatility under different market conditions. Energy Econ.
Guo, P., Zhu, H., You, W., 2018. Asymmetric dependence between economic policy 104851.
uncertainty and stock market returns in G7 and BRIC: a quantile regression Rehman, M.U., Apergis, N., 2019. Sensitivity of economic policy uncertainty to investor
approach. Finance Res. Lett. 25, 251–258. sentiment. Stud. Econ. Finance.
Guo, H., Hung, C.-H.D., Kontonikas, A., 2020. Investor sentiment and the Pre-FOMC Scarcioffolo, R.A., Etienne, X.L., 2018. Does Economic Policy Uncertainty Affect Energy
announcement drift. Finance Res. Lett. 101443. [Link] Market Volatility and Vice-Versa? Selected Paper Prepared for Presentation at the
frl.2020.101443. 2018 Agricultural Applied Economics Association Annual Meeting, Washington, D.
Hailemariam, A., Smyth, R., Zhang, X., 2019. Oil prices and economic policy uncertainty: C., August 5-August 7.
evidence from a nonparametric panel data model. Energy Econ. 83, 40–51. Salisu, A., Adediran, I., 2020. Uncertainty due to infectious diseases and energy market
Handley, K., Limão, N., 2017. Policy uncertainty, trade, and welfare: theory and evidence volatility. Energy Research Letters 1 (2), 14185.
for China and the United States. Am. Econ. Rev. 107 (9), 2731–2783. Sharif, A., Aloui, C., Yarovaya, L., 2020. COVID-19 pandemic, oil prices, stock market,
Howrey, E.P., 2001. The predictive power of the index of consumer sentiment. Brookings geopolitical risk and policy uncertainty nexus in the US economy: fresh evidence
Pap. Econ. Activ. (1), 175–207, 2001. from the wavelet-based approach. Int. Rev. Financ. Anal. 101496.
Huang, J., Li, Y., Zhang, H., Chen, J., 2021. The effects of uncertainty measures on Smales, L.A., 2019. Geopolitical Risk and Volatility Spillovers in Oil and Stock Markets.
commodity prices from a time-varying perspective. Int. Rev. Econ. Finance 71, SSRN 3414134.
100–114. Song, M., Shin, K.S., 2019. Forecasting economic indicators using a consumer sentiment
Ikonen, P., Kerola, E., Vilmi, L., 2019. The trade war has significantly weakened the index: survey-based versus text-based data. J. Forecast. 38 (6), 504–518.
global economy. Bank Finland Bull. 93 (4), 35–42. Steinberg, J.B., 2019. Brexit and the macroeconomic impact of trade policy uncertainty.
Kang, W., Ratti, R.A., 2013. Structural oil price shocks and policy uncertainty. Econ. J. Int. Econ. 117, 175–195.
Modell. 35, 314–319. Su, C.W., Khan, K., Tao, R., Nicoleta-Claudia, M., 2019. Does geopolitical risk strengthen
Karabulut, G., Bilgin, M.H., Doker, A.C., 2020. The relationship between commodity or depress oil prices and financial liquidity? Evidence from Saudi Arabia. Energy
prices and world trade uncertainty. Econ. Anal. Pol. 66, 276–281. 187, 116003.
Kilian, L., 2009. Not all oil price shocks are alike: disentangling demand and supply Wang, W., Yang, F., 2018. The shale revolution, geopolitical risk, and oil price volatility.
shocks in the crude oil market. Am. Econ. Rev. 99, 1053–1069. Geopolitical Risk, and Oil Price Volatility (August 30, 2018).
Kilian, L., Park, C., 2009. The impact of oil price shocks on the US stock market. Int. Wang, E.-Z., Lee, C.-C., 2020. Dynamic spillovers and connectedness between oil returns
Econ. Rev. 50, 1267–1287. [Link] and policy uncertainty. Appl. Econ. 52 (35), 3788–3808.
Koop, G., Korobilis, D., 2014. A new index of financial conditions. Eur. Econ. Rev. 71, Wei, Y., Liu, J., Lai, X., Hu, Y., 2017. Which determinant is the most informative in
101–116. forecasting crude oil market volatility: fundamental, speculation, or uncertainty?
Korobilis, D., Yilmaz, K., 2018. Measuring dynamic connectedness with large bayesian Energy Econ. 68, 141–150.
VAR models. WP, SSRN. [Link] Wen, F., Zhao, Y., Zhang, M., Hu, C., 2019. Forecasting realized volatility of crude oil
Kusek, P., Saurav, A., Kuo, R., 2020. Outlook and Priorities for Foreign Investors in futures with equity market uncertainty. Appl. Econ. 51 (59), 6411–6427.
Developing Countries: Findings from the 2019 Global Investment Competitiveness Yin, L., 2016. Does oil price respond to macroeconomic uncertainty? New evidence.
Survey in 10 Middle-Income Countries. Global Investment Competitiveness Report. Empir. Econ. 51 (3), 921–938.
World Bank Group, pp. 24–54. Yang, L., 2019. Connectedness of economic policy uncertainty and oil price shocks in a
Li, B., Chang, C.P., Chu, Y., Sui, B., 2020a. Oil prices and geopolitical risks: what time domain perspective. Energy Econ. 80, 219–233.
implications are offered via multi-domain investigations? Energy Environ. 31 (3), You, W., Guo, Y., Zhu, H., Tang, Y., 2017. Oil price shocks, economic policy uncertainty
492–516. and industry stock returns in China: asymmetric effects with quantile regression.
Li, F., Huang, Z., Zhong, J., Albitar, K., 2020b. Do tense geopolitical factors drive crude Energy Econ. 68, 1–18.
oil prices? Energies 13, 4277. Yao, J., Partington, G., Stevenson, M., 2012. Predicting the directional change in
Li, X.L., Balcilar, M., Gupta, R., Chang, T., 2016. The causal relationship between consumer sentiment. Aust. J. Manag. 38 (1), 67–80.
economic policy uncertainty and stock returns in China and India: evidence from a Youssef, M., Mokni, K., Ajmi, A.N., 2021. Dynamic connectedness between stock markets
bootstrap rolling window approach. Emerg. Mark. Finance Trade 52 (3), 674–689. in the presence of COVID-19 pandemic: does economic policy uncertainty matter?
Liang, C., Wei, Y., Li, X., Zhang, X., Zhang, Y., 2020. Uncertainty and crude oil market Financial Innovation forthcoming.
volatility: new evidence. Appl. Econ. 52 (27), 2945–2959. Youssef, M., Mokni, K., 2020. Modeling the relationship between oil and USD exchange
Liu, J., Ma, F., Tang, Y., Zhang, Y., 2019. Geopolitical risk and oil volatility: a new rates: evidence from a regime-switching-quantile regression approach. J. Multinatl.
insight. Energy Econ. 104548. Financ. Manag. 100625.
Liu, T., Gong, X., 2020. Analyzing time-varying volatility spillovers between the crude oil Zhang, B., 2019. Economic policy uncertainty and investor sentiment: linear and
markets using a new method. Energy Econ. 87, 104711, 2020. nonlinear causality analysis. Appl. Econ. Lett. 1–5.

19