Resources Policy 72 (2021) 102110

Contents lists available at ScienceDirect

Resources Policy
journal homepage: http://www.elsevier.com/locate/resourpol

On the connection between oil and global foreign exchange markets: The
role of economic policy uncertainty
Ismail O. Fasanya a, *, Oluwasegun B. Adekoya c, Abiodun M. Adetokunbo b
a
School of Economics and Finance, University of the Witwatersrand, Johannesburg, South Africa
b
Department of Economics, Augustine University, Epe, Lagos, Nigeria
c
Department of Economics, Federal University of Agriculture, Abeokuta, Ogun State, Nigeria

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

Keywords: This paper examines the effect of U.S. economic policy uncertainty on the connectedness across oil and the most
Economic policy uncertainty globally traded currency pairs. First, we examine volatility spillover among oil and the exchange rates, and find
Oil-exchange rates nexus strong connection between crude oil and currency markets with oil being net receivers of shocks. Second, BDS
Nonparametric quantile causality
test shows that nonlinearity is very important when examining the role of EPU in affecting the interactions
Volatility spillover
between oil and exchange rate markets. Third, the nonparametric quantiles-based causality test shows that the
spillover for each asset is driven by economic policy uncertainty around the lower and median quantiles. This
finally suggests that the role of the U.S. economic policy in influencing global financial cycle which consequently
leads to capital flows and movements in the prices of assets across financial markets cannot be overemphasized.
Relevant policy implications can be drawn from these findings.

1. Introduction consequently its output price, relative to another economy whose

tradable sector is not energy intensive. This leads to the currency
As the global economy stands at present, the crude oil and foreign appreciation of the energy-dependent economy (see Buetzer et al., 2016;
exchange markets rank high in the general categories of commodity and Chen and Chen, 2007). Krugman (1983) introduces the wealth effect
financial markets respectively. The importance of these markets is mechanism by arguing from the standpoint of oil-exporting countries. As
strictly connected to their facilitation of global economic activities. For oil price rises, current account balance of these countries also improves
instance, crude oil is adjudged not only as constituting the largest share in local currency terms following an increase in oil revenue. The end
of global energy use (Atems et al., 2015), but also the most globally result is an expectation of the appreciation of their currencies, while
traded commodity (Demirer et al., 2020). On the other hand, due to the those of the oil dependent countries depreciate (see Bechmann and
increasing level of globalization, international dependency and varia­ Czudaj, 2013). Beckmann et al. (2020) further argues that this scenario
tion in countries’ currencies, the foreign exchange market is highly also has the tendency to cause short-run appreciation of the U.S. dollar if
indispensable in aiding economic activities at the global scale. Mean­ the huge revenues of the oil exporters are reinvested in assets that are
while, foreign transactions involving crude oil is denominated in the priced in the U.S. dollar. However, the wealth effect is generally effec­
world’s most prominent currency, U.S. dollar, (see Jain and Biswal, tive in the short-run (see Beckmann et al., 2020). The medium-run and
2016). This implies that domestic currencies of countries must be con­ long-run implication of the scenario is termed the portfolio effect, and it
verted to the U.S. dollar to trade in the global crude oil market. This relates particularly to the U.S. dollar in terms of the currencies of
serves as the first basis upon which both markets are linked. oil-exporting countries. Considering first a case where the U.S. is a net
The literature has thus established various mechanisms through oil exporter, trade effects will either depreciate the U.S. dollar against
which the crude oil and foreign exchange markets are connected. The the currencies of the oil-exporters if the demand for oil by the U.S.
terms of trade channel due to Amano and Van Norden (1998a,b) comes dominates, or appreciate it if the demand for U.S. commodities domi­
from the perspective of relative price changes. Increase in oil price raises nates (Beckmann et al., 2020). The other aspect of the portfolio channel
the aggregate price level of an energy-dependent economy and is in terms of the desire of the oil exporting countries for medium- or

* Corresponding author.
E-mail addresses: [email protected] (I.O. Fasanya), [email protected] (O.B. Adekoya), [email protected] (A.M. Adetokunbo).

https://doi.org/10.1016/j.resourpol.2021.102110
Received 6 August 2020; Received in revised form 13 February 2021; Accepted 8 April 2021
Available online 23 April 2021
0301-4207/© 2021 Elsevier Ltd. All rights reserved.
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

long-term financial assets denominated in the U.S. dollar. This leads to (2012) to examine how the uncertainty connects the spillovers in the
the demand for U.S. dollar thus causing it to appreciate (Buetzer et al., crude oil and foreign exchange markets. Financial and high frequency
2016; Coudert et al., 2008). series are mostly known to exhibit heavy tails, excess kurtosis and
The literature also notes that there is a possibility for exchange rate non-normality. These are often as a result of inherent nonlinearity,
to affect oil price. This theoretical proposition is led by the currency- structural breaks and regime changes. In the presence of these, linear
pricing basis of oil price which is the U.S. dollar. Assuming that oil is frameworks become inappropriate. This therefore motivates our choice
homogenous and internationally traded, the denomination channel of the nonlinear test. This approach has two noble merits as highlighted
holds that the depreciation of the U.S. dollar causes a fall in domestic in Balcilar et al. (2015). The first advantage is that it produces reliable
currency-based oil price for the foreign countries. Oil demand by these results in the presence of functional misspecification errors and depen­
countries then fall, and the eventual outcome is a reduction in the U.S. dence of series, such as is common to financial series. Its other benefit is
dollar-based currency (see Bloomberg and Harris, 1995; Akram, 2009). the dual causality testing. Apart from testing for causality in conditional
The last theoretical proposition between the two markets discloses that mean, it additionally provides results for causality in conditional vari­
oil and foreign exchange markets are linked through common indicators ance of series due to the non-normal distribution property of most
including gross domestic products (GDP), stock prices, interest rates, etc. financial series. Obviously, the nonlinear causality tests explored in
These factors jointly affect the two markets thereby linking them. Albulescu et al. (2019) cannot capture the nonlinear-causality in
So far, the literature is fraught with empirical studies that examine conditional-variance. Also, the study was unable to reveal the level at
the link between the two markets with varying findings resulting. The which causality holds unlike the quantiles-based approach used in this
first class of studies examines causal relationship either in a unidirec­ study. Meanwhile, the validity of the quantiles-based causality test is
tional or bidirectional framework (see Beckmann and Czudaj, 2013; first established by applying another test, namely BDS test developed by
Benassy-Quere et al., 2007; Coudert et al., 2008, Chen and Chen, 2007, Brock et al. (1996) to check for the evidence of nonlinearity in the series.
etc.), while another focuses on the impact of shocks especially from oil This is perhaps another contribution of this study rather than a mere
price (see Buetzer et al., 2016; Huang and Guo, 2007; Basher et al., speculation approach employed in most past studies.
2012). Due to various economic crisis and policy interventions, the The remainder of the paper is given the following structure. Next
relationship between oil price and exchange rate has also been found to section outlays a brief review of the related literature. Methodological
be nonlinear and time-varying (see Akram, 2004; Zhang, 2013; Basher approaches adopted in this study is elaborated in Section 3 with a
et al., 2016, Fan and Xu, 2011; Zhang et al., 2008; etc.). description of data. In Section 4, we offer a detailed discussion on the
In line with the objective of our study, we also observe that the empirical analysis and the last section offers conclusion of our study.
literature on dynamic spillover among the crude oil and various foreign
exchange markets is growing recently (see Malik and Umar, 2019; 2. Brief review of literature
Albulescu et al., 2019; Tiwari and Albulescu, 2016; Jain and Biswal,
2016). However, what is yet to be explored is how the spillover among The literature is replete with studies on the links among oil, foreign
them is driven by notable exogenous factors, especially economic or exchange markets and economic policy uncertainty, each producing
financial markets uncertainty. The earliest theoretical foundation for unique results following their employment of variety of techniques for
this consideration is rooted in the works of Pastor and Veronesi (2012) different economies. Essentially, oil-exchange rates connectedness could
and Gomes et al. (2012), followed by the subsequent studies of Arouri predict as well as respond to other economic and financial series (Xu
et al. (2016), Liu et al. (2017), etc. They note that returns and volatility et al., 2019; Dai et al., 2020).
connectedness in financial markets can be influenced by uncertainty Earlier, we have highlighted the various mechanisms through which
through its impacts on the supply of labour, personal consumption and oil and foreign exchange markets are connected. Based on these estab­
investment decisions. From another perspective, policy-induced uncer­ lished mechanisms, the relationship between both markets has been
tainty has diverse impacts on corporate entities, investors and con­ examined in different forms. There is a strand of the literature that be­
sumers as it can dissuade corporations from involving in new investment lieves that the connection between the markets is dependent, among
projects and induce conservative spending behaviour of consumers (see other factors, on the nature of oil price shocks and underlying nature of
Converse, 2017; Handley and Limao, 2015). Similarly, higher economic the economy under consideration. For instance, one of the standing
policy uncertainty causes lenders to be conservative in their lending empirical discoveries is that demand- and risk-driven oil price shocks
habits which then causes interest rates to rise. It is thus not out of scope significantly contribute to variation in exchange rates (Malik and Umar,
to assert that the wide effect of economic policy uncertainty on the 2019; Xu et al., 2019), while impacts from oil supply-side shocks are
economy can directly creep into financial and oil markets (see Albulescu insignificant (see Xu et al., 2019). This partly contrasts the findings of
et al., 2019). Jiang et al. (2020) which disclose that oil supply and oil specific demand
However, empirical studies have largely linked policy uncertainty shocks have negative and asymmetric effects on exchange rate, and that
with financial and oil markets in different scenes, with majority focusing developed currencies exhibit significant Granger-causality relationship
on stock markets. Even the recent studies (see, for, instance, Badshah coming from oil shocks. Huang et al. (2020) also observe that unex­
et al., 2018; Fang et al., 2018) are limited in scope as they only link pected oil price shocks could have greater influence on currency markets
policy uncertainty with stock and oil markets. It thus still remains a over time. However, Wen and Wang (2020) prove that oil exports and
debatable issue on how economic policy uncertainty drives the volatility foreign exchange regimes are the important factors that determine the
spillovers across the foreign exchange and oil markets. The only known volatility transmission across foreign exchange markets. On the other
work till date that mirrors ours is Albulescu et al. (2019) which connects hand, the conditions of currency and oil markets explain the responses of
the spillover between oil and commodity currency markets with the U.S. exchange rates to oil price shocks (Youssef and Mokni, 2020).
economic policy uncertainty. However, our present study differs from Another strand of the literature seems to be concerned with the time-
theirs in three distinct ways. First, we determine how the spillovers varying and nonlinear properties of the oil-foreign exchange markets
among the most traded foreign exchanges and oil market are connected nexus. Beckmann et al. (2020) confirm the links between oil prices and
to policy-based uncertainty, rather than commodity currency and the U. exchange rates to be time-varying and volatile. In support of nonlinear
S. economic policy-based uncertainty. We opine that, following the causality, Tiwari et al. (2019) show that the nature and direction of the
degree of integration among the countries whose foreign exchange causality differs from one country to another among the BRICS econo­
markets are being studies, the exchange rates would be more sensitive to mies. Corroborating the nonlinear evidence, Mensi et al. (2017) reveal
uncertainty in economic policy since they are the most traded. the presence of asymmetric systemic risks from oil to currencies, and at
Secondly, we use a quantile-based causality test of Jeong et al. the same time, from currencies to oil. Using time-varying tri-variate

2
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

vine-copula quantile regression model, Dai et al. (2020) prove that the attempt to address this deficiency include Xu et al. (2019) and Dai et al.
US foreign exchange market connects oil and gold market in the (2020) which respectively consider business cycle and gold as factors
short-run time scales, caused probably due to the fact that the two that connect both the oil and foreign exchange markets. This study ad­
commodities are largely denominated in the U.S. dollar. Although, Chkir dresses these concerns by: (i) looking into the dynamic connectedness
et al. (2020) find oil to be a weak hedge against exchange rates, Ahmad between oil and globally traded exchange rates; (ii) determining the role
et al. (2020) discovers that oil and currency markets shocks have im­ of the U.S. economic policy uncertainty following the impact of the
mediate reverse transmission mechanism. The latter further supports the country in inducing global credits and capital flows. For now, only the
asymmetric impact of currency price appreciation, or otherwise, on the study of Albulescu et al. (2019) resembles ours, but we consider globally
oil market. traded exchange rates rather than commodity currencies used by them.
More recently, empirical focus has been driven to the spillover Furthermore, our methodological choice is superior as it accounts for
relationship between the oil and foreign exchange markets, either causality in both conditional mean and conditional variance.
exclusively or in addition to other international financial markets.
Albulescu et al. (2019) examine the time- and frequency-domain spill­ 3. Methodology and data
over between oil and commodity currencies, and how the connectedness
is driven by economic policy uncertainty. They report that the financial 3.1. The Diebold – Yilmaz spillover approach
markets are strongly connected with oil averagely being a net trans­
mitter of shocks to the currencies across all the frequency cycles. More This study uses the Diebold and Yilmaz (DY, 2012) framework to
particularly, the currencies of New Zealand and Australia are the most examine the connection between oil and foreign exchange rate markets.
sensitive to shocks in the oil shocks, explained by their use in carry trade The DY framework for the spillover analysis is grounded on the forecast
strategies globally. This is in line with the findings of Singh et al. (2018) error variance decomposition from the generalized VAR framework of
that the general currency market is a net receiver of shocks from the oil Koop et al. (1996) and Pesaran and Shin (1998), hereafter KPPS, which
market. Furthermore, Malik and Umar (2019) provide evidence to produces variance decompositions which are invariant to the ordering.
support the intensified degree of connectedness between oil price shocks In setting up the spillover indexes, a covariance stationary VAR (ρ) is
and exchange rate since the outbreak of financial crisis. The very recent considered.1
study of Adekoya and Oliyide (2020) also discover that the oil and U.S. ( ∑)
currency markets, in addition to other financial markets, are strongly rt = Φrt− 1 + εt , εt ∼ ∼ 0, (1)
connected during the COVID-19 pandemic period.
Meanwhile, since the construction of the economic policy uncer­ where rt = (r1t, r2t,......, rNt ) is an N × 1 vector of volatility series, Φ is an
tainty index by Baker et al. (2016), there has been tremendous influx of N × N matrix of parameters, εt is a vector of independently and identi­
∑
studies on the impact of economic policy uncertainty on various eco­ cally distributed disturbances and is the variance matrix for the error
nomic and financial indicators. On the role of economic policy uncer­ vector ε . The moving average representation can be written as:
tainty on foreign exchange rate market, Kido (2016) establishes the ∑∞
spillover effect of the U. S. economic policy uncertainty on real effective rt = i=0
Ai εt− 1 (2)
exchange rates. Accounting for nonlinearity and structural breaks,
Al-Yahyaee et al. (2020) also confirms economic policy uncertainty and where Ai is assumed to obey the recursion Ai = Φ1 Ai− 1 + Φ2 Ai− 2 + ... +
currency markets are nonlinearly related, as also consistent with the Φp Ai− p . A0 is an identity matrix with an N × N dimension and Ai = 0 for
discovery of Chen et al. (2020). i < 0. Equation (2) forms the basis for the derivation of variance de­
Some studies take another dimension to examine the response of compositions required to determine the spillover indexes. The spillover
exchange rate to economic policy uncertainty. For instance, economic involves an own and cross variances share, where the former is defined
policy uncertainty intensifies exchange rate volatility as disclosed by as the fractions of the H-step-ahead error variances in forecasting ri that
Krol (2014), Bartsch (2019) and Zhou et al. (2020), and improves the are due to shocks to ri , for i = 1, 2....N and the latter is the fractions of
forecasting power of macroeconomic models of exchange rates in the H-step ahead error variances in forecasting ri that are due to shocks
different horizons (Abid, 2020). Beckmann and Czudaj (2017) find that to rj , for i, j = 1, 2, ...., N for such that i ∕
= j.
exchange rate expectations are affected by uncertainty regarding future Based on the generalized VAR framework of KPPS, H -step-ahead
g
position of economic policy. Though, Huynh et al. (2020) examined and forecast error variance decompositions denoted by θij is written as:
found strong volatility connectedness between foreign exchange rates
∑1 (
H− ∑ )2
and policy uncertainty, this study does not examine the role of policy σ−jj 1
′
e i Ah ej
uncertainty on the connectedness of oil and exchange rates markets, θgij (H) = h=0
(3)
which is the objective of this study. ∑1(
H−
′ ∑ ′
)
e i Ah Ah e i
Turning to the link between economic policy uncertainty and the h=0

crude oil market, Hailemariam et al. (2019) reveals nonlinear and

Where σjj is the standard deviation of ε for the jth equation and ei is the
time-varying relationship. In support of this, Yang (2019) establishes
connectedness and causal relationship between economic policy un­ selection vector, with one as the ith element and zeros otherwise. Since
certainty and oil price shocks. They additionally confirm that crude oil the sum of the contributions to the variance of the forecast error is not
∑ g
prices receive information from economic policy uncertainty. Fang et al. equal to one – that is Nj=i θij (H) ∕
= 1 ; DY (2012) normalized each entry
(2018) also reveal the role of economic policy uncertainty in the cor­ of the variance decomposition matrix by the row sum in order to use the
relation between oil and stock markets. full information of the matrix. The normalized KPPS H -step-ahead
g
So far, it can be seen that there is a substantial number of studies on forecast error variance decompositions represented by ̃ θ (H) is ij
the oil-exchange rate nexus, as well as the role of economic policy un­ expressed as:
certainty. However, there are still some crucial gaps yet to be addressed
by the previous studies. First, most studies on the oil-exchange rate
nexus are in the form of causal impact, with less investigation on their
dynamic connectedness. Second, the role of economic policy uncertainty
on oil market and foreign exchange rate markets has only been
considered independently. The determinants of the connectedness be­ 1
See Diebold and Yilmaz (2012) paper for a detailed exposition of the
tween both markets are yet to be significantly explored. Studies that methodology.

3
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

g θgij (H) (EPU) does not cause yt (market spillovers) in the σ − quantile with
̃
θij (H) = ∑N g (4) respect to the lag-vector of {yt− 1 , …, yt− q , xt− 1 , xt− q } if
j=1 θij (N)
( ⃒ ) ( )
∑N ̃g ∑N ̃g Qσ yt ⃒yt− 1 , …, yt− q , xt− 1 , …, xt− q = Qσ yt |yt− 1 , …, yt− q (10)
where j=1 θij (H) = 1 and 1,j=1 θij (H) = N by construction.
Given these preliminaries, the total spillover index is written as: While xt causes yt in the σ th quantile with respect to {yt− 1 , …, yt− q ,
xt− 1 , xt− q } if
∑N ̃g i, j = 1 ∑N ̃g i, j = 1 ( ⃒ ) ( )
θij (H)
i =
∕ j
θij (H)
i∕=j Qσ yt ⃒yt− 1 , …, yt− q , xt− 1 , xt− q ∕= Qσ yt |yt− 1 , …, yt− q (11)
Sg (H) = ∑N g × 100 = × 100 (5)
̃ N
θ
1,j=1 ij (H) Definitively, Qσ (yt | ⋅) = σth quantile of yt depending on t and 0 < σ <
All the parameters in equation (5) have been previously defined. 1. We denote Vt− 1 ≡ (yt− 1 , …,yt− q ), Ut− 1 ≡ (xt− 1 , …, xt− q ), and Wt = (Ut ,
⃒ ⃒
Essentially, equation (5) measures the contribution of spillovers of Vt ); and Fyt |Wt− 1 (yt ⃒Wt− 1 ) and Fyt |Vt− 1 (yt ⃒Vt− 1 ) represents the conditional
⃒
volatility shocks across the assets under consideration. In our case, the distribution of yt given Wt− 1 and Vt− 1 , respectively. Also, Fyt |Vt− 1 (yt ⃒Vt− 1 )
total spillover index captures the contribution of spillovers of volatility
is assumed to be absolutely continuous in yt for almost all Wt− 1 . If we
shocks across the six (6) currency pairs and oil price to the total forecast
proceed by denoting Qσ (Wt− 1 ) ≡ Qσ (yt |Wt− 1 ) and Qσ (Vt− 1 ) ≡ Qσ (yt |Vt− 1 ),
error variance.
then we have Fyt |Wt− 1 {Qσ (yt |Wt− 1 )} = σ with a probability of one. In
Also, it is possible to assess quantitatively the direction of spillovers
across the entire markets using the DY (2012) approach. These direc­ essence, the hypothesis to be tested based on the specified definitions in
tional spillovers are classified into two namely ‘Directional Spillover To’ equations (10) and (11) are
and ‘Directional Spillover From’. The former measures the directional { }
H0 = P Fyt |Wt− 1 {Qσ (yt |Wt− 1 )} = σ = 1, (12)
spillovers whether volatility transmitted by market i to all other markets
j while the latter relates to volatility received by market i from all other { }
H1 = P Fyt |Wt− 1 {Qσ (yt |Wt− 1 )} = σ < 1, (13)
markets j. The index for the computation of ‘Directional Spillover To’
g
denoted by S.i is given as: Furthermore, Jeong et al. (2012) utilize the distance measure J =
{τt E(τt |Wt− 1 )fW (Wt− 1 )}, where τt and fz (Wt− 1 ) are the regression error
∑ ∑ and marginal density function of Zt− 1 , respectively. The regression error
N g N g
̃
θij (H) ̃
θij (H)
i, j = 1 i, j = 1 emanates through its basis in the null hypothesis as specified in equation
i∕
=j i∕
=j (12), which can only be true if and only if E[1{yt ≤ Qσ (Vt− 1 )|Wt− 1 )} = σ
S.ig (H) = × 100 = × 100 (6) or, equivalently, 1{yt ≤ Qσ (Vt− 1 )} = σ + τt , where 1{ ⋅} is the indicator
∑
N g N
̃
θji (H) function. Thus, Jeong et al. (2012) specify the distance measure, G ≥ 0,
i,j=1
as follows:
Also, the ‘Directional Spillover From’ denoted as Sgi. is measured using [{ }2 ]
G = E Fyt |Wt− 1 {Qσ (yt |Wt− 1 )} − σ fW (Wt− 1 ) (14)
the index given below:
∑
N g ∑
N g It is pertinent to note that, we will have a situation where G = 0 if
̃
θij (H) ̃
θij (H) and only if the null in equation (12) is true, while we will have G > 0
i, j = 1 i, j = 1
under the alternative hypothesis in equation (13). Also, Jeong et al.
i∕
=j i∕
=j (2012) introduced a feasible kernel-based test statistic for J which has
Si.g (H) = × 100 = × 100 (7)
∑
N
̃g N the following form:
θij (H)
i,j=1
∑
T ∑
T ( )
1 Wt− 1 − Zs−
(15)
̂T = 1
Equally, the Net Spillovers can be obtained using the index expressed G K τ̂t τ̂s ,
T(T − 1)s2q t=q+1 r=q+1,r∕
=t
s
below:
Where K( ⋅) denotes the kernel function with bandwith s. T, q, τ̂t is
Sig (H) = S.ig (H) − Si.g (H). (8)
the sample size, lag-order and estimate of the regression error, respec­
Equation (8) gives the difference between the gross volatility shocks tively. The estimate of the regression error is computed as thus:
transmitted to and received from all other markers. In other words, in­ { }
formation about each market’s contribution to the volatility of other
̂σ (Yt− 1 ) − σ
τ̂t = 1 yt ≤ Q (16)
markers can be obtained through the net spillovers.
Also, we further use the nonparametric kernel method to estimate
To examine the net pairwise volatility spillover between markets i − 1
and j, we compute the difference between the gross volatility shocks the σth conditional quantile of yt given Vt− 1 as Q F yt |Vt− 1 (σ |Vt− 1 ),
̂ σ (Vt− 1 ) = ̂
transmitted from market i to market j and those transmitted from j to i: where the Nadarya-Watson Kernel estimator is specified as follows
⎡ ⎤ ( ) ( )
[ g ] ∑T Vt− 1 − Vr− 1
⎢ ̃ g g
θij (H) ⎥
̃ ̃
θji (H) − ̃
g N 1 y r ≤ y t
⎢ θji (H) ⎥ θij (H) r=q+1,r∕
=t s
g
Sij (H) = ⎢ N − N ⎥.100 = .100 (9) ̂ yt |Vt− 1 (yt |Vt− 1 ) =
F ( ) (17)
⎣ ∑ ̃g ∑ ̃g ⎦ N ∑T Vt− 1 − Vr− 1
θik (H) θjk (H) =t N
r=q+1,r∕ s
i,k=1 j,k=1

In the analysis of this paper, the second order 7-variable VARs with Where N( ⋅) is the kernel function and s is the bandwidth.
10-step-ahead forecasts was considered. By extension, Balcilar et al. (2018) extends the framework of Jeong
et al. (2012) by developing a test for the second moment. Thus, they
adopt the nonparametric Granger-quantile-causality approach by Nish­
3.2. Nonlinear causality test
iyama et al. (2011). To illustrate the causality in higher order moment,
we assume
The study adopts the Balcilar et al. (2018) methodology which is
crucial for the detection of nonlinear causality via a hybrid approach yt = h(Vt− 1 ) + ϑ(Ut− 1 )τt , (18)
with a foundation from the frameworks of Nishiayama et al. (2011) and
Jeong et al. (2012). As noted by Jeong et al. (2012), the variable xt Where τt is the white noise process and h( ⋅) and ϑ( ⋅) equals the unknown

4
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

functions that satisfy pertinent conditions for stationarity. Although, this of oil worldwide, and is used to price two thirds of the world’s inter­
specification allows no granger-type causality testing from Ut− 1 to yt , nationally traded crude oil supplies. The data is derived from the FRED
however, it could detect the “predictive power” from Ut− 1 to y2t when database of the Federal Reserve Bank of St. Louis source from the
ϑ( ⋅) is a general nonlinear function. Thus, the study re-formulate DataStream database of Thomson Reuters. For the proxy of the EPU, we
equation (18) to account for the null and alternative hypothesis for select the US EPU index constructed by Baker et al. (2016), which are
causality in variance in equations (19) and (20), respectively. available for download from Baker’s personal website, http://www.po
{ } licyuncertainty.com.
H0 = P Fy2 |Wt− 1 {Qσ (yt |Wt− 1 )} = σ = 1, (19) For the basic conditions of stationarity of the variables required for
t
our nonlinear causality to hold, we decided to work with logarithmic
{ }
H1 = P Fy2 |Wt− 1 {Qσ (yt |Wt− 1 )} = σ < 1, (20) returns series of both oil price and the currency pairs (daily natural
t
logarithmic change expressed in percentages) since both series returns
This study obtains the feasible test statistic for the testing of the null were non-stationary following the standard unit root tests.3 As for the
hypothesis in equation (19), and then replace yt in equation (15) – (17) EPU, we work with the logarithmic levels of EPU index which was later
with y2t (that is, volatility). With the inclusion of Jeong et al. (2012) found to be stationary following the standard unit root tests.4
approach, the study overcomes the issue that causality in mean implies
causality in variance. Specifically, the study interprets the causality in 4. Empirical results
higher-order moments through the use of the following model:
4.1. Preliminary analyses results
yt = h(Ut− 1 , Vt− 1 ) + τt , (21)
Thus, we specify the higher order quantile causality as It is often a standard practice in most empirical studies that the
{ } prerequisite knowledge about the statistical properties of the underlying
H0 = P Fyk |Wt− 1 {Qσ (yt |Wt− 1 )} = σ = 1, for k = 1, 2, …, k, (22) series be first provided. We keep this norm by revealing brief statistical
information about the series. The descriptive statistics reported in
t

{ } Table 1 shows that oil price and all the major exchange rates considered
H1 = P Fyk |Wt− 1 {Qσ (yt |Wt− 1 )} = σ < 1, for k = 1, 2, …, k. (23)
t except USD/CHF, GBP/USD and USD/CAD observe positive returns on
Overall, we test that xt granger causes yt in σth quantile up to the K-th average over the period under consideration, indicating a likely evi­
moment through the use of equation (22) to construct the test statistic of dence of positive gains by investors. However, the negative returns
equation (15) for each k. Although, Nishiyama et al. (2011) note that it recorded by the three highlighted exchange rates do not indicate losses
is not easy to combine different statistics for each k = 1, 2, …, k into one at all times. In fact, among the exchange rates, USD/CHF and GBP/USD
statistic for the joint null in equation (22) which is mutually correlated. whose average returns are negative give the highest positive returns
However, to circumvent this issue, we adopt a sequential-testing method (9.1950% and 8.7246% respectively). In terms of their volatilities, we
as described by Nishiyama et al. (2011) with some modifications. To observe that with the standard deviation value of oil price which is
begin with, we test for the nonparametric granger causality in mean 2.1811, it is more than twice as volatile as the most volatile exchange
(k = 1). Failure to reject the null of k = 1 does not translate into non rate (AUD/USD) whose standard deviation statistic is 0.7893. USD/CAD
causality in variance, thus, we construct the tests for k = 2. Finally, we exhibits the least degree of variability. It is not surprising how that the
test for the existence of causality-in-mean and variance successively. We Jarque-Bera test rejects the null hypothesis of normal distribution for all
determine the lag order using SIC. The bandwidth is selected through the series following from the reports of both the skewness and kurtosis
the use of least squares cross-validation method. For K( ⋅) and L( ⋅), we statistics. While the skewness values hover between positive and nega­
utilize the Gaussian kernels. tive for all the returns series, their kurtosis estimates are exceedingly
larger than the standard threshold. This suggests the presence of extreme
fluctuations in these financial and commodity markets. Interestingly,
3.3. Data description Albulescu et al. (2019) find evidence largely similar to this study for
their commodity currencies and WTI oil price.
This paper utilizes daily data of oil price, six (6) currency pairs and Certain implications could be drawn from this brief descriptive
Economic Policy Uncertainty (EPU) index of the US covering the period analysis. Firstly, the non-normality of the series gives a relative indica­
between 1/4/2000 and 6/25/2020. The start and end dates are gov­ tion of heavy right or left tail and excess kurtosis. This could further
erned based on data availability of US EPU index. The currency pairs suggest the presence of nonlinearity and/or structural shifts along the
considered for the analysis are The Australian Dollar and US Dollar time paths of the series such that the use of linear or constant parameter
(AUD/USD) (nicknamed ‘aussie’), The British Pound and US Dollar models would bring about spurious results (see Adekoya and Oliyide,
(GBP/USD) (nicknamed ‘cable’), The Euro and US Dollar (EUR/USD) 2020b; Adekoya et al., 2020). This gives a concrete justification for our
(nicknamed ‘euro’), The US Dollar and Canadian Dollar (USD/CAD) choice of quantiles-based causality test. Secondly, the evidence of heavy
(nicknamed ‘loonie’), The US Dollar and Japanese Yen (USD/JPY) tails as well as high volatility passes motivates the necessity to examine
(nicknamed ‘gropher’), and The US Dollar and Swiss Franc (USD/CHF) the relationship in both the conditional-mean and conditional-variance
(nicknamed ‘swissie’). These currency pairs, along with their various (see Balcilar et al., 2015).
combinations (such as EUR/JPY, GBP/JPY and EUR/GBP) have
continued to drive all speculative trading in global FX.2These currency 4.2. Spillover results
pairs are the most traded currency pairs in the world that account for
more than 95% of all speculative trading in global Foreign exchange Since the objective of this study is to examine how economic policy
(Salisu et al., 2018). The currency pairs data are freely downloadable
from the database of Forex Forum Global View (www.global-view.
com/forex-trading-tools/forexhistory/). As for oil price, we use the 3
The full details of the preliminary analysis are available on request from the
daily price of Brent Crude as it serves as a benchmark price for purchases authors.
4
These results contradict the theoretical argument of measures of uncertainty
which are meant to be stationary. However, the statistical results presented
2
This statement is ascribed to Boris Schlossberg at: http://www.investo here deviate from this which may be as a result of the sample frame used in this
pedia.com/articles/forex/06/sevenfxfaqs.asp. study. The full preliminary results are available on request from the authors.

5
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

Table 1
Descriptive statistics of returns series.
Exchange rates and oil price Mean Maximum Minimum Std. dev. Skewness Kurtosis Jarque-Bera

EUR/USD 0.0017 3.6763 − 2.8181 0.6005 0.0321 4.8744 783.1127

USD/JPY 0.0010 3.8320 − 4.5829 0.6168 − 0.1989 6.8387 3315.7605
USD/CHF − 0.0094 9.1950 − 15.7117 0.6791 − 1.9438 64.5226 846007.90
GBP/USD − 0.0053 8.7246 − 8.3235 0.6018 − 0.5536 25.7768 115766.36
USD/CAD − 0.0011 3.2737 − 4.3346 0.5438 0.1031 6.5802 2862.9915
AUD/USD 0.0007 5.6326 − 10.0619 0.7893 − 0.7951 14.2468 28722.872
Oil price 0.0090 32.3825 − 40.7835 2.1811 − 2.0643 68.5346 959919.51

EUR/USD, USD/JPY, USD/CHF, GBP/USD, USD/CAD, and AUD/USD respectively represent euro to U.S. dollar, U.S. dollar to Japanese Yen, U.S. dollar to Swiss Franc,
Pounds to U.S. dollar, U.S. dollar to Canadian dollar and Australian dollar to U.S. dollar. The bolded values of the Jarque-Bera statistic indicate the rejection of the null
hypothesis of normal distribution of series.

uncertainty (EPU) drives the connectedness among oil and financial rate channel) that discloses the possibility of oil price to be driven by
markets, it is expedient to first reveal the dynamic volatility spillover exchange rates (see Akram, 2009).
among the markets. The summarized spillover results are shown in Concluding this section on volatility spillovers in the financial and oil
Table 2. Uniquely, the table shows the contribution to the variance markets, it is evident that there is established transmission of shocks.
forecast errors of a particular asset to and from other assets. Then, the Although the degree and direction of shocks transmission vary, just like
net spillover is computed by the difference between the total contribu­ oil seems to be closely knitted to USD/CAD and AUD/USD, and USD/
tions given by an asset and the total it gives, with a positive value CHF being a weak receiver of shocks, the overall performance still
implying that the asset in question is a net transmitter, rather than a net suggests significant connectedness among the markets. Also, three fac­
receiver. In other words, if the net spillover value is positive, then the tors seem to be obvious for the transmission of shocks in the markets.
asset transmits more shocks or information to other assets than it re­ The first is the degree of integration among the G-7 countries whose
ceives from them. We observe from Table 2 that the highest receiver of exchange rates are the most traded and being considered in this study.
shocks from all the remaining assets combined is USD/CAD with a value The integration results in high transmission of shocks among them. The
of 36.10, followed by AUD/USD whose value is 27.20. Although the second factor is the role of Canada and Australia in the global supply of
shocks received by the oil market from all the major currencies appear crude oil. This, coupled with the fact that oil is denominated in the U.S.
small (2.70), but it is still fair compared to USD/CHF that is not dollar, leads the strong volatility spillover connection between oil and
responsive to shocks from even its co-currencies. Interestingly, USD/ each of USD/CAD and AUD/USD. Thirdly, the Australian dollar, as
CAD and AUD/USD are still among the highest transmitters of shocks, in revealed by Albulescu et al. (2019) is often associated with high rates of
addition to EUR/USD and USD/CHF. We could link the likely reason for interest compared to other currencies, thus making it to enjoy wide use
the high spillovers of USD/CAD and AUD/CAD to the significant role of as an investment currency when making trade strategies. In other words,
Canada and Australia in the global supply of the most globally traded investors would shun currencies with lower interest rates in favour of
commodity, i.e. crude oil. The high degree of integration of the G-7 the ones with high interest rates. Again, U.S. and Canada are closely
countries is another factor that drives high transmission of shocks related in terms of economic cooperation and trade agreements, and the
among them. former is one of the largest importers of the crude oil of the latter.
Furthermore, it is seen that among the currency pairs, EUR/USD and Therefore, the combination of all these factors may be responsible for
USD/CAD receive more shocks than they give, thus leading to their the connectedness between the currency and oil markets.
negative net spillover values of − 5.10 and − 17.30 respectively. Oil is Linking these spillover transmissions to uncertainty in economic
also a net receiver of shocks (− 1.5) being inconsistent with Albulescu policy, the U.S. is an indispensable factor driving the global financial
et al. (2019) and Sing et al. (2018). The reason is obviously due to dif­ cycle through her various monetary policies pronounced by the Federal
ferences in the currency pairs considered. However, oil seems to connect Reserves. Thus, the connectedness across the markets may be driven by
stronger with USD/CAD and AUD/USD (see Appendix Table A) than policy uncertainty having first affected global liquidity and investors’
other currency pairs, again confirming the role of Canada and Australia decisions. This implies that uncertainty in economic policy that drives
in the global oil market, and reflecting the U.S. dollar as the major fluctuations in exchange rate and/or oil price may induce volatility
currency for pricing crude oil. Thus, this is probably the reason oil is a shocks to the other markets. The possibility of economic policy uncer­
net receiver of shocks in line with the theoretical proposition (exchange tainty to affect the volatility spillover between the currency and oil
markets is therefore the main thrust of this paper and focused on in the
next section.
Table 2
Spillover results.
4.3. Causality test results
Exchange Total contribution Net
rates and oil spillover
From To To self Including Following the observed evidence of volatility transmissions between
price (b-a)
others others (c) own (b + c) the foreign exchange and oil markets, we proceed to the examination of
(a) (b) the causal effect of economic policy uncertainty on the established
EUR/USD 26.80 21.70 73.40 95.10 − 5.10 connectedness in the markets. Doing this requires that we test the null
USD/JPY 24.40 31.10 75.70 106.80 6.70 hypothesis that economic policy uncertainty does not cause the total
USD/CHF 0.00 0.10 100.00 100.10 0.10
spillover and the net spillover for each currency pair and oil price under
GBP/USD 6.90 8.60 93.10 101.70 1.70
USD/CAD 36.10 18.80 63.80 82.60 − 17.30 consideration. We initially examine the causal effect from the perspec­
AUD/USD 27.20 43.60 72.80 116.2 16.40 tive of linear relationship with the results reported in Table 3. It is
Oil price 2.70 0.20 97.30 97.50 − 1.5 observed that the effect of economic policy uncertainty is found to be
EUR/USD, USD/JPY, USD/CHF, GBP/USD, USD/CAD, and AUD/USD respec­ insignificant in most cases. The three exemptions are the total spillover,
tively represent euro to U.S. dollar, U.S. dollar to Japanese Yen, U.S. dollar to net spillovers for USD/JPY and oil price, which are all significant at
Swiss Franc, Pounds to U.S. dollar, U.S. dollar to Canadian dollar and Australian 10%.
dollar to U.S. dollar. However, we perceive that this weak performance of the policy-

6
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

Table 3 out that although some studies favour graphical presentation of the re­
Linear causality test results. sults (see, for instance, Balcilar, et al., 2018; Balcilar et al., 2015), this
Null hypothesis F- Prob. value study considers the tabular presentation in order to reveal the depth of
statistics the nonlinear relationship at all the conventional significance levels, i.e.
EPU does not Granger-cause Total spillover 2.6845* 0.0684 1%, 5% and 10%. The limitation of the graphical presentation is that,
EPU does not Granger-cause Net spillover for EUR/USD 0.2946 0.7448 significance line is drawn for only 5% significance level, thus making
EPU does not Granger-cause Net spillover for USD/JPY 2.3748* 0.0932 significance at 10% to be missed. Notwithstanding, we still present the
EPU does not Granger-cause Net spillover for USD/CHF 0.7831 0.4571 quantiles-based graphical results in Appendix Figs. 1 and 2. In sharp
EPU does not Granger-cause Net spillover for GBP/USD 0.5747 0.5629
EPU does not Granger-cause Net spillover for USD/CAD 0.4299 0.6506
contrast to the results of the linear causality test, Tables 5 and 6 which
EPU does not Granger-cause Net spillover for AUD/USD 0.7337 0.4802 reports the nonlinear results for the conditional-mean and
EPU does not Granger-cause Net spillover for oil 2.4386* 0.0874 conditional-variance show strong evidence of the rejection of the null
* denotes significance at 10% critical level.
hypothesis of no Granger-causality. The causal evidence is mostly sig­
nificant at the lower quantiles, with some reaching the median region.
However, the causality becomes weak at the extreme quantiles, sug­
based uncertainty in affecting the connectedness in the markets is due to
gesting that the effect of economic policy on the connectedness among
likely presence of nonlinearity in the series. At the most basic level, the
the markets is sensitive to the degree of the performance of the foreign
presence of heavy tails, excess kurtosis and non-normality are pointers
exchange and oil markets. When the markets are performing at their
to the possibility of nonlinear nature of the series. However, we conduct
peak, economic policy seems to be weak in affecting their interactions.
a more formal test (BDS test)5 developed by Brock et al. (1996) to
In summary, our results reveal three facts: (i) there is strong
establish the presence of nonlinearity in the series. The BDS test results
connectedness between the foreign exchange and oil markets; (ii) the
(see Table 4) show strong evidence of nonlinear relationship between
connectedness among these markets are primarily driven by economic
economic policy uncertainty and all the spillover series as the null hy­
policy uncertainty, although the causal effect seems to be stronger
pothesis of serial dependence is resoundingly rejected across all di­
around the lower and middle quantiles in most cases; (iii) nonlinearity is
mensions. An implication of these results is that there is more to what
a very crucial factor to be put into consideration when examining the
the linear Granger-causality test reveals, it likely could have suffered
role of economic policy uncertainty in affecting the interactions between
from the problem of misspecification.
foreign exchange and oil markets. In these scenarios, our results confirm
Having detected strong evidence of nonlinear relationship in the
those of Albulescu et al. (2019) who reveal that commodity currencies
relationship between economic policy uncertainty and the connected­
and oil market are dynamically connected, and policy-induced uncer­
ness among the assets, we turn to the results of the quantiles-based
tainty is significant in driving this interaction. Fortunately, their
causality test. In order not to miss out from any important informa­
nonlinear causality techniques are different from the one explored in
tion, the quantiles-based causality analysis is conducted in both the
this study, but yet, the results do not differ. This indicates that the
conditional-mean and conditional-variance. It is also important to point
impact of economic policy uncertainty on the interactions among
financial and commodity markets is stable and strong. On the other
hand, although their study is mainly on the impact of economic policy
Table 4
BDS test of independence results. uncertainty on stock returns, Balcilar et al. (2015) use similar technique
as this study (quantiles-based causality test) to prove that the jettisoning
Spillovers 2 3 4 5 6
nonlinearity in the predictability of financial variables (and their
Total 0.0175*** 0.0316*** 0.0385*** 0.0395*** 0.0373*** connectedness) with economic policy uncertainty may lead to unreliable
spillover
results.
Net spillover 0.0178*** 0.0339*** 0.0417*** 0.0436*** 0.0425***
for EUR/
USD 5. Conclusion
Net spillover 0.0201*** 0.0354*** 0.0424*** 0.0447*** 0.0437***
for USD/
Following the recent development and computation of the index of
JPY
Net spillover 0.0180*** 0.0339*** 0.0412*** 0.0430*** 0.0421***
economic policy uncertainty by Baker et al. (2016), an ongoing move in
for USD/ the literature relates to how international markets are connected by this
CHF new measure of uncertainty. The validity of this recent empirical drive
Net spillover 0.0161*** 0.0298*** 0.0383*** 0.0403*** 0.0386*** also has to do with the high degree of integration being observed in these
for GBP/
markets, thus leaving a puzzle as to what could be the likely cause. Could
USD
Net spillover 0.0163*** 0.0309*** 0.0399*** 0.0431*** 0.0431*** uncertainty induced by economic policy be responsible? In sharp
for USD/ contrast to most past empirical studies that merely examined the impact
CAD of economic policy uncertainty on the performance of financial markets
Net spillover 0.0150*** 0.0298*** 0.0362*** 0.0380*** 0.0368***
in terms of their volatility and returns (see for instance Liu et al., 2017;
for AUD/
USD
Kang et al., 2017; You et al., 2017; Arouri et al., 2016; etc.), this study
Net spillover 0.0229*** 0.0410*** 0.0501*** 0.0530*** 0.0512*** focus on how the U.S. economic policy uncertainty affect connectedness
for oil among oil and the globally most traded currency pairs. The significant
The reported values are the BDS statistics. *** indicates significance at 1% role of the U.S. in driving global financial cycle through capital flows
critical level. and movements in the prices of assets across financial and commodity
markets make us consider the uncertainty resulting from her economic
policy. Thus, our analyses are in two folds, with the first being the
evaluation of volatility spillover among the considered assets (oil and
the exchange rates). The second entails the assessment of the effect of
5
The BDS test is carried out on the residuals of each spillover series (overall the economic policy uncertainty on the total spillover and the net
and net) in the VAR(1) model that includes the EPU. In other words, the EPU spillover series of each asset through the use of the quantiles-based
index and each of the spillover series are captured in a VAR(1) model, after causality test which handles inherent nonlinearity, structural breaks
which the residuals of the latter are generated. Then, the BDS test is conducted and regime changes in the series.
on the generated residuals (see Balcilar et al., 2015 for a similar approach). Expectedly, our findings show that there is a significant

7
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

Fig. 1. Non-linear causality in mean for net spillovers.

Fig. 2. Non-linear causality in variance for net spillover.

Table 5
Quantile-based (nonlinear) causality test in conditional mean results.
Quantiles Null hypothesis: EPU does not cause:

Total EUR/USD USD/JPY USD/CHF GBP/USD USD/CAD AUD/USD Oil price

0.1 1.8713* 3.4716*** 3.3027*** 4.1172*** 1.9591* 2.8091*** 2.4981** 3.2796***

0.2 2.2077** 2.1173** 1.9816** 2.4350** 1.8330* 1.5729 1.8146* 2.3962**
0.3 2.9792*** 1.3841 1.4443 1.6680* 1.7159* 1.6879* 1.8874* 2.4423**
0.4 2.7578*** 1.3174 1.0954 1.5759 2.6863*** 1.9240* 1.5424 1.7186*
0.5 2.3346** 1.1762 0.8866 1.1897* 2.0811** 1.5953 1.3895 1.2431
0.6 2.6183*** 0.9419 0.8596 0.9007 1.5072 1.2036 1.1705 0.8128
0.7 2.3584** 0.8411 0.7844 0.5211 1.0445 0.7767 0.7980 0.4905
0.8 1.5891 0.6106 0.6518 0.2680 0.8768 0.5916 0.6029 0.2516
0.9 0.9391 1.2883 1.3101 0.0914 0.3480 0.5132 0.5977 0.0853

Values reported are t-statistics. ***, ** and * respectively denote significance at 1%, 5% and 10% critical levels.

connectedness among the assets, with EUR/USD, USD/CAD and oil mean and conditional variance, the economic policy uncertainty
being net receivers of shocks, and others net transmitters. With this, we significantly affects the net spillover of each asset, mostly at the lower
are well positioned to conduct the causality test. However, we first and middle quantiles. In essence, we find, in line with the global
conduct the linear Granger-causality test with the results revealing weak financial cycle channel, that risk transmissions across the oil and most
evidence of causal effect of economic policy uncertainty on the volatility globally traded currency pairs are driven by uncertainty induced by the
connectedness measures. This we believe is as a result of the nonline­ U.S. economic policy.
arity features of the series, which is further confirmed by the BDS test. Accordingly, these findings present opportunities for viable policy
The quantile-based nonlinear results show that, for both conditional implications for both investors/portfolio managers and policy

8
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

Table 6
Quantile-based (nonlinear) causality test in conditional variance results.
Quantiles Null hypothesis: EPU does not cause:

Total EUR/USD USD/JPY USD/CHF GBP/USD USD/CAD AUD/USD Oil price

0.1 2.2116** 3.0664*** 3.2834*** 3.7829*** 1.9864** 1.4104 4.0854*** 2.6653***

0.2 2.2734** 2.2400** 1.8956* 2.1963** 2.0097** 1.5439 2.4269** 2.3182**
0.3 3.0161*** 1.5827 1.2734 1.4817 1.9856** 1.3255 2.2216** 1.5990
0.4 2.7178*** 1.4089 1.1985 1.0882 1.9685** 1.3535 1.7237* 1.5225
0.5 2.3363** 1.1685 1.0712 0.8783 1.3048 1.5892 1.3629 1.1580
0.6 2.4536** 0.8464 0.7565 0.7530 1.6493* 1.3517 1.0819 0.8430
0.7 2.2824** 0.6371 0.5647 0.5211 1.0311 1.0717 0.8292 0.6450
0.8 1.7253* 0.2608 0.2769 0.6064 0.5358 0.6386 0.3382 0.2516
0.9 0.7604 0.0892 0.0950 0.0914 0.2933 0.2459 0.1168 0.0853

Values reported are t-statistics. ***, ** and * respectively denote significance at 1%, 5% and 10% critical levels.

implications. For the former, the weak connectedness of oil price and the firms, while policy makers can find it difficult to implement appropriate
USD/CHF and GBP/USD exchange rates with other assets in the system macroeconomic policies required to optimally manage the economy’s
suggest that these assets can assets can serve as safe haven during pe­ finances to enhance economic growth and development and other
riods of extreme market stress. This is particularly stronger for the USD/ macroeconomic objectives. Finally, policy makers must closely monitor
CHF which receives no shocks from other assets. Turning to the latter, movement in the uncertainty of the U.S. economic policies as it is a
policy makers need to introduce policies that safeguard the risks expo­ major driver of the connectedness between the crude oil and exchange
sures of both the crude oil and foreign exchange markets. This is rates.
because, for the crude oil market, risks exposure could result into sig­
nificant fluctuations in oil price which would eventually affect both the Author statement
oil exporters and oil importers adversely, depending on the direction of
the movement in the oil price. On the other hand, with the currency The authors of this article certify that we have seen and approved the
pairs being the most globally traded, their undue exposure to risks can final version of the manuscript being submitted. It is our original work,
lead to significant fluctuations and consequently impose harmful effects hasn’t received prior publication and isn’t under consideration for
for multinational corporations, investors and macroeconomic perfor­ publication elsewhere. There are no conflicts of interests among the
mance. Investments in foreign assets denominated in these currencies authors.
can result into significant losses for investors and the multinational

Appendix
Table A1
Full spillover results

Exchange rates and oil price From Total spillover from others Net spillover

Exchange rates Oil price

EUR/USD USD/JPY USD/CHF GBP/USD USD/CAD AUD/USD

To EUR/USD 73.40 6.20 0.10 4.30 7.80 8.40 0.00 26.80 − 5.10
USD/JPY 6.70 75.70 0.00 2.60 2.10 13.00 0.00 24.40 6.70
USD/CHF 0.00 0.00 100.00 0.00 0.00 0.00 0.00 0.00 0.10
GBP/USD 1.90 2.60 0.00 93.10 1.30 1.10 0.00 6.90 1.70
USD/CAD 7.30 6.80 0.00 0.90 63.80 21.00 0.10 36.10 − 17.30
AUD/USD 5.80 13.10 0.00 0.80 7.40 72.80 0.10 27.20 16.40
Oil price 0.00 2.40 0.00 0.00 0.20 0.10 97.30 2.70 − 1.5
Total spillover to others 21.70 31.10 0.10 8.60 18.80 43.60 0.20 124.10

References Albulescu, C.T., Demirer, R., Raheem, I.D., Tiwari, A.K., 2019. Does the U.S. economic
policy uncertainty connect financial markets? Evidence from oil and commodity
currencies. Energy Econ. 83, 375–388.
Adekoya, O.B., Oliyide, J.A., 2020. How COVID-19 drives connectedness among
Al-Yahyaee, K.H., Shahzad, S.J.H., Mensi, W., 2020. Tail dependence structures between
commodity and financial markets: evidence from TVP-VAR and causality-in-
economic policy uncertainty and foreign exchange markets: nonparametric quantiles
quantiles techniques. Resour. Pol. https://doi.org/10.1016/j.
methods. International Economics 161 (c), 66–82.
resourpol.2020.101898.
Amano, R., Van Norden, S., 1998a. Exchange rates and oil prices. Rev. Int. Econ. 6,
Adekoya, O.B., Oliyide, J.A., Oduyemi, G.O., 2020. How COVID-19 upturns the hedging
683–694.
potentials of gold against oil and stock markets risk: nonlinear evidence through
Amano, R., Van Norden, S., 1998b. Oil prices and the rise and fall of the US real exchange
threshold regression and Markov regime switching models. Resour. Pol. https://doi.
rate. J. Int. Money Finance 17, 299–316.
org/10.1016/j.resourpol.2020.101926.
Arouri, M., Estay, C., Rault, C., Roubaud, D., 2016. Economic policy uncertainty and
Abid, A., 2020. Economic policy uncertainty and exchange rates in emerging markets:
stock markets: long-run evidence from the U.S. Financ. Res. Lett. 18, 136–141.
short and long run evidence. Finance Res. Lett. 37.
Atems, B., Kapper, D., Lam, E., 2015. Do exchange rates respond asymmetrically to
Ahmad, W., Prakash, R., Uddin, G.S., Chahal, R.J.K., Rahman, M.L., Dutta, A., 2020. On
Shocks in the Crude oil market? Energy Econ. 49 (c), 227–238.
the intraday dynamics of oil price and exchange rate: what can we learn from China
Badshah, I., Demirer, R., Suleman, T., 2018. The effect of economic policy uncertainty on
and India? Energy Econ. 91.
stock-commodity correlations and its implications on optimal hedging. Work. Pap.
Akram, Q.F., 2004. Oil prices and exchange rates. Norwegian evidence. Econ. J. 7,
https://doi.org/10.2139/ssrn.3240962. Available at: SSRN 3240962.
476–505.
Baker, S., Bloom, N., Davis, S., 2016. Measuring economic policy uncertainty. Q. J. Econ.
Akram, Q.F., 2009. Commodity prices, interest rates and the dollar. Energy Econ. 31,
131, 1593–1636.
838–851.

9
I.O. Fasanya et al. Resources Policy 72 (2021) 102110

Balcilar, M., Gupta, R., Kyei, C., 2015. South African Stock Returns Predictability Using Kang, W., Perez de Gracia, F., Ratti, R.A., 2017. Oil price shocks, policy uncertainty, and
Domestic and Global Economic Policy Uncertainty: Evidence from a Nonparametric stock returns of oil and gas corporations. J. Int. Money Finance 70, 344–359.
Causality-In-Quantiles Approach. University of Pretoria Department of Economics Kido, Y., 2016. On the link between the US economic policy uncertainty and exchange
Working Paper Series. Working Paper: 2015-70. rates. Econ. Lett. 144, 49–52.
Balcilar, M., Gupta, R., Pierdzioch, C., Wohar, M.E., 2018. Terror attacks and stock Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in non-linear
market fluctuations: evidence based on a nonparametric causality-in-quantiles test multivariate models. J. Econom. 74, 119–147.
for the G7 countries. Eur. J. Finance 24 (4), 333–346. Krol, R., 2014. Economic policy uncertainty and exchange rate volatility. Int. Finance 17
Bartsch, Z., 2019. Economic policy uncertainty and dollar-pound exchange rate return (2), 241–256.
volatility. J. Int. Money Finance 98. Krugman, P., 1983. Oil and the dollar. In: Jagdeeps, B., Buldfordh, P. (Eds.), Economic
Basher, S.A., Haug, A.A., Sadorsky, P., 2012. Oil prices, exchange rates and emerging Interdependence and Flexible Exchange Rates. MIT Press, Cambridge, MA.
stock markets. Energy Econ. 34, 227–240. Liu, Z., Ye, Y., Ma, F., 2017. Can economic policy uncertainty help to forecast the
Basher, S.A., Haug, A.A., Sadorsky, P., 2016. The impact of oil shocks on exchange rates: volatility: a multifractal perspective. Physica A 482, 181–188.
a Markov-switching approach. Energy Econ. 54, 11–23. Malik, F., Umar, Z., 2019. Dynamic connectedness of oil price shocks and exchange rates.
Beckmann, J., Czudaj, R., 2017. Exchange rate expectations and economic policy Energy Econ. 84 (c).
uncertainty. Eur. J. Polit. Econ. 47 (c), 148–162. Mensi, W., Hammoudeh, S., Shahzad, S.J.H., Al-Yahyaee, K.H., Shahbaz, M., 2017. Oil
Beckmann, J., Czudaj, R., 2013. Oil prices and effective dollar exchange rates. Int. Rev. and foreign exchange market tail dependence and risk spillovers for MENA,
Econ. Finance 27, 621–636. emerging and developed countries: VMD decomposition based copulas. Energy Econ.
Beckmann, J., Czudaj, R.L., Arora, V., 2020. The relationship between oil prices and 67, 476–495.
exchange rates: revisiting theory and evidence. Energy Econ. https://doi.org/ Nishiyama, Y., Hitomi, K., Kawasaki, Y., Jeong, K., 2011. A consistent nonparametric test
10.1016/j.eneco.2020.104772. for nonlinear causality - specification in time series regression. J. Econom. 165,
Benassy-Quere, A., Mignon, V., Penot, A., 2007. China and the relationship between the 112–127.
oil price and the dollar. Energy Pol. 35, 5795–5805. Pastor, L., Veronesi, P., 2012. Uncertainty about government policy and stock prices.
Bloomberg, S.B., Harris, E.S., 1995. The commodity-consumer price connection: fact or J. Finance 67, 1219–1264.
fable? Federal Reserve Board of New York. Economic Policy Review 21–38. Pesaran, M.H., Shin, Y., 1998. Generalized impulse response analysis in linear
Brock, W., Dechert, D., Scheinkman, J., LeBaron, B., 1996. A test for independence based multivariate models. Econ Lett 1998 58, 17–29.
on the correlation dimension. Econom. Rev. 15, 197–235. Salisu, A.A., Oyewole, O.J., Fasanya, I.O., 2018. Modelling return and volatility
Buetzer, S., Habib, M.M., Stracca, L., 2016. Global exchange rate fluctuations: do oil spillovers in global foreign exchange markets. J. Inf. Optim. Sci. 39 (7), 1417–1448.
shocks matter? IMF Econ. Rev. 64, 443–470. Singh, Vipul Kumar, Nishant, Shreyank, Kumar, Pawan, 2018. Dynamic and directional
Chen, L., Du, Z., Hu, Z., 2020. Impact of economic policy uncertainty on exchange rtae network connectedness of crude oil and currencies: evidence from implied volatility.
volatility of China. Finance Res. Lett. 32. Energy Econ., Elsevier 76 (C), 48–63.
Chen, S.-S., Chen, H.-C., 2007. Oil price and real exchange rates. Energy Econ. 29, Tiwari, A.K., Albulescu, C.T., 2016. Oil price and exchange rate in India: fresh evidence
390–404. from continuous wavelet approach and asymmetric, multi-horizon Granger-causality
Chkir, I., Guesmi, K., Brayek, A.B., Naoui, K., 2020. Modelling the nonlinear relationship tests. Appl. Energy 179, 272–283.
between oil prices, stock markets, and exchange rates in oil-exporting and oil- Tiwari, A.K., Trabelsi, N., Alqahtani, F., Bachmeier, L., 2019. Modelling systemic risk and
importing countries. Res. Int. Bus. Finance 54. dependence structure between the prices of crude oil and exchange rates in BRICS
Converse, N., 2017. Uncertainty, capital flows, and maturity mismatch. J. Int. Money economies: evidence using quantile coherency and NGCoVaR approaches. Energy
Finance 88, 260–275. Econ. 81 (c), 1011–1028.
Coudert, V., Mignon, V., Penot, A., 2008. Oil price and the dollar. Energy Stud. Rev. 15, Wen, T., Wang, G.-J., 2020. Volatility connectedness in global foreign exchange markets.
45–58. J. Multinatl. Financ. Manag. 54.
Dai, X., Wang, Q., Zha, D., Zhou, D., 2020. Multi-scale dependence structure and risk Xu, Y., Han, L., Wan, L., Yin, L., 2019. Dynamic link between oil prices and exchange
contagion between oil, gold, and US exchange rate: a wavelet-based vine-copula rates: a nonlinear approach. Energy Econ. 84 (c).
approach. Energy Econ. 88 (c). You, W., Guo, Y., Zhu, H., Tang, Y., 2017. Oil price shocks, economic policy uncertainty
Demirer, R., Ferrer, R., Shahzad, S.J.H., 2020. Oil price shocks, global financial markets and industry stock returns in China: asymmetric effects with quantile regression.
and their connectedness. Energy Econ. 88. Energy Econ 68 (2017), 1–18.
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional Yang, L., 2019. Connectedness of economic policy uncertainty and oil price shocks in a
measurement of volatility spillovers. Int. J. Forecast. 28, 57–66. time domain perspective. Energy Econ. 80 (c), 219–233.
Fan, Y., Xu, J.-H., 2011. What has driven prices since 2000? A structural change Youssef, M., Mokni, K., 2020. Modeling the relationship between oil and USD exchange
perspective. Energy Econ. 33, 1082–1094. rates: evidence from a regime-switching-quantile regression approach. J. Multinatl.
Fang, L., Chen, B., Yu, H., Xiong, C., 2018. The effect of economic policy uncertainty on Financ. Manag. 55 (c).
the long-run correlation between crude oil and the U.S. stock markets. Finance Res. Zhang, Y., 2013. The link between the price of oil and the value of the US dollar. Int. J.
Lett. 24, 56–63. Energy Econ. Pol. 3 (4), 341–351.
Gomes, F.J., Kotlikoff, L.J., Viceira, L.M., 2012. The excess burden of government Zhang, Y.J., Fan, Y., Tsai, H.T., Wei, Y.M., 2008. Spillover effect of US dollar exchange
indecision. Tax Pol. Econ. 26, 125–164. rate on oil prices. J. Pol. Model. 30, 973–991.
Hailemariam, A., Smyth, R., Zhang, X., 2019. Oil prices and economic policy uncertainty: Zhou, Z., Fu, Z., Jiang, Y., Zeng, X., Lin, L., 2020. Can economic policy uncertainty
evidence from a nonparametric pane; data model. Energy Econ. 83, 40–51. predict exchange rate volatility? New evidence from GARCH-MIDAS model. Finance
Handley, K., Limao, N., 2015. Trade and investment under policy uncertainty: theory and Res. Lett. 34.
firm evidence. Am. Econ. Pol. 7, 189–222.
Huang, S., An, H., Lucey, B., 2020. How do dynamic responses of exchange rates to oil
price shocks co-move? From a time-varying perspective. Energy Econ. 86 (c). Further reading
Huang, Y., Guo, F., 2007. The role of oil price shocks on China’s real exchange rate.
China Econ. Rev. 18 (4), 403–416. Alam, M.S., Shahzad, S.J.H., Ferrer, R., 2019. Causal flows between oil and forex markets
Huynh, T.L.D., Nasir, M.A., Nguyen, D.K., 2020. Spillovers and connectedness in foreign using high-frequency data: asymmetries from good and bad volatility. Energy Econ.
exchange markets: the role of policy uncertainty. The Quarterly Review of https://doi.org/10.1016/j.eneco.2019.104513.
Economics and Finance. https://doi.org/10.1016/j.qref.2020.09.001. Baghestani, H., Chazi, A., Khallaf, A., 2019. A directional analysis of oil prices and real
Jain, A., Biswal, P.C., 2016. Dynamic linkages among oil price, gold price, exchange rate exchange rates in BRIC countries. Res. Int. Bus. Finance. https://doi.org/10.1016/j.
and stock market in India. Resour. Pol. 49, 179–185. ribaf.2019.06.013.
Jeong, K., Härdle, W.K., Song, S., 2012. A consistent nonparametric test for causality in Bai, S., Koong, K.S., 2017. Oil prices, Stock returns and exchange rates: empirical
quantile. Econ. Theor. 28, 861–887. evidence from China and the United States. N. Am. J. Econ. Finance. https://doi.org/
Jiang, Y., Feng, Q., Mo, B., Nie, H., 2020. Visiting the effects of oil price shocks on 10.1016/j.najef.2017.10.013.
exchange rates: quantile-on-quantile and Causality-in-quantiles approaches. N. Am. Cifarelli, G., Paladino, G., 2010. Oil price dynamics and speculation: a multivariate
J. Econ. Finance 52. financial approach. Energy Econ. 32, 363–372.

10