Economic Change and Restructuring (2023) 56:4603–4628

https://doi.org/10.1007/s10644-023-09565-x

Oil supply and oil price determination among OPEC

and non‑OPEC countries: Bayesian Granger network
analysis

David Oluseun Olayungbo1 · Aziza Zhuparova2 ·

Mamdouh Abdulaziz Saleh Al‑Faryan3,4

Received: 26 October 2022 / Accepted: 3 October 2023 / Published online: 4 November 2023
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature
2023

Abstract
This study examines the relationship between oil production and oil price deter-
mination among OPEC and non-OPEC countries during the period of 1965–2021.
The study period chosen captures perfectly the various historical geopolitical crises,
oil shock supply, GFCS and recent global health pandemic that have stirred up the
global issue of oil price war among leading oil producing countries. After carrying
out the preliminary tests, we adopt both the static and dynamic Bayesian Graphi-
cal Network causality along with the traditional Granger causality test for robust-
ness analysis. Our results show Saudi Arabia as the world leading producer of oil
and Russia as the main determinants of the global oil prices, while USA acts as a
stabilizer in the global oil market. The reliability of our results lies in the very low
in-sample forecast error. The implication of our results is that Saudi Arabia, Rus-
sia and USA as leading oil producing countries matter for oil price and oil supply
determination in the global market. Therefore, the three countries should cooperate
with other OPEC and non-OPEC member countries than compete to ensure stabil-
ity in oil price and the global economy. The outcome of this study can serve as a
guide to oil producing cartels in coordinating the global oil price for global eco-
nomic stability.

* David Oluseun Olayungbo

[email protected]
1
Obafemi Awolowo University, Ile‑Ife, Nigeria
2
Higher School of Economics and Business, AL‑Farabi, Kazakh National University, Almaty,
Kazakhstan
3
School of Accounting, Economics and Finance, Faculty of Business and Law, University
of Portsmouth, Portsmouth, UK
4
Consultant in Economics and Finance, Riyadh, Saudi Arabia

13
Vol.:(0123456789)
4604 Economic Change and Restructuring (2023) 56:4603–4628

Keywords Oil supply · Oil price war · OPEC · US · Russia · Bayesian · MCMC ·
Granger causality

JEL Classification G12 · E32 · Q4

1 Introduction

Apart from the oil demand factors, supply factors have also become major deter-
minants of the global oil price movement. Supply factors such as technological
innovations in oil production, Organization of Petroleum Exporting Countries’
(OPEC) decisions, global financial crises (GFCS), health pandemic crisis and so
on have become fundamentals to oil price variations. According to Kilian (2009),
oil price movement are affected differently depending on either the oil shock
movement is demand or supply driven. The global oil supply disruption can be
historically traced to many past crises such as the Arabian embargo in 1973,
Iranian revolution in 1978, Iraq–Iran war in 1980–1988, Iraq attack of Kuwait
in 1990, Saudi Arabian Gulf war in 1990, Iraq war in 2003 and Venezuela civil
unrest in 2002 till 2003 (Kilian 2009). In addition, there was the oil supply dis-
ruption of United States Shale oil revolution in 2016 with oil price falling from
112 United State Dollar (USD) per barrel to 48 USD per barrel (Umechukwu
and Olayungbo 2022). The Russian–Ukraine conflict in 2014 is another oil supply
disruption that led to fall in the global oil price in that period (Dreger et al. 2015).
In April 2020, at the peak of the COVID-19 outbreak when the oil price was
at its lowest ebb, West Texas Intermediate (WTI) oil price dropped so much that
it became negative. Specifically, in 21 April 2020, the oil price of WTI in United
State Dollar (USD) per barrel was −37.63 and that of Brent oil price in USD
per barrel was 19.33, (Statista 2022). The negative oil price implies oil supply
was below the cost price which resulted in revenue and welfare loss of oil pro-
ducing countries. The drastic fall in the oil price due to COVID-19 lockdown
necessitated the agreement between Saudi Arabia and Russia to reduce oil supply
for the stabilization of global oil price at that time (Ma et al. 2021). However,
the OPEC + agreement could not hold. Price war resulted with Russia’s refusal
to cut oil production in spite of fall in oil price and Saudi Arabia responded with
increase in oil supply in April 2020 precisely (Singh 2020). The consequential
effect of the lack of agreement between the two major oil producing countries,
coupled with the COVID-19 pandemic, was a fall in oil prices and loss of welfare
nearly for all oil producing countries (Ma et al. 2021). Prior to 2020, similar price
war occurred in 2014, when OPEC decided at Vienna meeting to cut oil supply.
The move plummeted oil price against most oil producing countries at that year
such that huge loss of revenue and poor economic condition were recorded by
many member countries of both OPEC and non-OPEC (Yang et al. 2022).
It should be noted that a reasonable high level of oil price is important for oil
producing countries to benefit and grow their economies. Higher oil price is equally
important to attract international investors and for them to get profitable returns

13
Economic Change and Restructuring (2023) 56:4603–4628 4605

on their investment capital (OPEC 2012). It is therefore crucial for oil producing
countries to cooperate and not compete to ensure a relatively stable high oil price.
This is because when oil price is low, economies of oil exporting countries are neg-
atively affected especially if such countries depend on oil export revenue to meet
their financial obligations. Also, lower oil prices may constraint many oil producing
countries to fulfil their fiscal responsibilities. For instance, in the year 2020, Saudi
Arabia’s fiscal budget execution was tied to 60 USD per barrel of oil export while
it was 30–40 USD per barrel for Russia (US-Saudi Business Council 2020; Con-
sumer News and Business Channel (CNBC) 2020). The budgetary target is not the
same for another OPEC member country like Nigeria. In the same year, the bench
mark for the budget was 57 USD per barrel (This day newspaper in Nigeria 2020).
In other words, any fall in the global oil price below the expected oil bench price
makes successful budget implementation difficult. The implication of the differences
in oil bench price for OPEC and non-OPEC member country is that both parties
must work together to achieve a favorable oil price for positive growth of mem-
ber countries. Therefore, the purpose of this study is to determine the oil export-
ing country that drives the global oil price and to identify the leading and lagging
countries in the global oil market. Unveiling the leading and lagging countries in
the international oil market is significant for appropriate decision-making in interna-
tional economic policy in terms of oil supply cut to stabilize global oil price by the
oil producing cartels and regional countries such as OPEC, non-OPEC, Organisa-
tion for Economic Cooperation and Development (OECD), Commonwealth of Inde-
pendent States (CIS), non-OECD and European Union (EU) countries. The outcome
of this study can assist each oil producing cartel in their decisions to coordinate and
unify oil policies of its members in response to oil price changes for global oil sup-
ply and economic stability of member countries.
The present geopolitical war, specifically the ongoing Russia–Ukraine war has
also become a threat to global oil supply with increase in oil price from 74.17 USD
per barrel to 117.25 USD per barrel in February 24 2022 following Russia invasion
of Ukraine. The higher oil price motivated the USA to release more barrels of oil
from her reserve into the global market and implored OPEC, OPEC + and other oil
producing countries to follow suit. The step was taken to lower price and reduce
Russia’s financial ability to wage war against Ukraine. Many previous studies have
examined the effects of OPEC and non-OPEC oil supply on global oil price. For
instance, Khalid (2016) adopted Engle and Granger, Johansen cointegration and
ARDL to examine the long run relationship between OPEC oil supply and each
member’s supply from both monthly and quarterly data from 1994 to 2014. The
study found no existence of cartel between member countries and OPEC. The cau-
sality results however showed oil price as a driver of OPEC oil production. Colgan
(2014) also concluded that OPEC could not control oil price during the period of
study. On the contrary, Bremond et al. (2012) found OPEC oil production to deter-
mine oil price during oil shock periods. However, earlier study by Smith (2005)
opined OPEC market structure to be in between a cartel and non-cooperative oli-
gopoly. Alhajji and Huettner (2000) reported Saudi Arabia as a dominant producer
with OPEC as a non-cartel producer. Branger et al. (2019) in another paper stud-
ied the macroeconomic effects of oil price shocks from OPEC’s monopoly power

13
4606 Economic Change and Restructuring (2023) 56:4603–4628

and found expansion in global real economic activity from OPEC markup oil supply
shocks. Ma et al. (2021) investigated the impact of oil price war between Russia and
Saudi Arabia during COVID-19 on the crude oil market using event study and found
negative crude oil prices to significantly affect WTI crude oil market for the period
of March 2020 to May 2020. Another recent paper by Cai et al. (2022) used SVAR
model and examined the transmission effects of OPEC and non-OPEC oil supply
cuts on macroeconomic variables and discovered unexpected non-OPEC oil supply
to be more pronounced on the Euro area. Kisswani et al. (2022) also investigated
impact of non-OPEC oil supply on OPEC oil production for the period of Janu-
ary 1993 to March 2020 using Quantile autoregressive distributed lags (QARDL)
and reported long run symmetric impact of non-OPEC production on OPEC pro-
duction. Lastly, Yang et al. (2022), employing panel ordinary least square (OLS),
investigated the effects of oil production on global oil price in OPEC and non-OPEC
member countries from the month of January 1995 to November 2015. The study
reported that OPEC oil supply had significant effect on the global oil price than non-
OPEC oil supply.
This present study is different from previous studies on global oil price war
among OPEC and non-OPEC countries in two ways. First, this study makes use
of long historical data of major oil exporting countries during major geopolitical
wars, oil supply shocks and GFCS with the inclusion of COVID-19 health pandemic
period in order to determine the country or group that drives the global oil price
among US, Russia, Saudi Arabia and OPEC member countries. The current previous
studies such as Branger et al (2019), Ma et al. (2021), Cai et al. (2022), Yang et al.
(2022) and Kisswanni et al. (2022) have made use of shorter data span that have not
captured past supply shocks that are important for the global oil price and oil sup-
ply relationships. It is our beliefs that identifying the driver(s) of the global oil price
from this study can possibly help to mitigate price war and improve the welfare of
both oil producing and oil importing countries for global market stability. Secondly,
we adopt Bayesian graphical network (BGN) analysis to determine the countries that
drive the global oil price and to identify the countries and group that are the lead
(leader) and lag (follower) in the international oil market space for oil supply policy
coordination. Past studies earlier stated have employed frequentist approach such
as QARDL, OLS and SVAR in their methodologies. We have, on the other hand,
adopted a Bayesian approach of graphical network. The BGN is deemed appropriate
for this kind of study from the structural view of causal network of interactions of
oil players such as OPEC and non-OPEC member countries employing oil supply
quota as a tool to determine oil prices in the global oil market. It is assumed that
such network of interactions is best modeled with the BGN method. The approach
involves the use of multivariate instantaneous and autoregressive causal structure
with posterior inference generated from Markov Chain Monte Carlo (MCMC) simu-
lation with Gibbs sampling. The BGN method is therefore advantageous over the
frequentist causal approach because of its efficient use of conditional probabilistic
prior for its posterior inference. With the BGN, we are able to estimate the static
(instantaneous) and dynamic (autoregressive) relationships between oil supply and
oil price determination among OPEC and non-OPEC in Bayesian network structure.

13
Economic Change and Restructuring (2023) 56:4603–4628 4607

To the best of our knowledge, this study is among the first to investigate this area of
economic research through the Bayesian approach.
In overall, the static and the dynamic results show that Saudi Arabia is the lead-
ing and dominant player in the global oil market by controlling the OPEC oil sup-
ply. The finding suggests policy coordination between OPEC members and Saudi
Arabia on oil production. Given the result, it is clear that the market structure of
OPEC is a cartel or Oligopolistic in nature. In addition, Russia and the US are found
as followers to Saudi Arabia in the international market with dynamic relationship
among them. Lastly, Russia oil supply solely determines the global oil prices of both
brent and WTI while the US acts as both the oil price and oil supply stabilizers.
This study therefore concludes that coordinated oil supply action is necessary for
all the major three players in the global oil market space to achieve stabilization of
the global oil market, in terms of sustainable oil supply to consumers and a steady
income to its member countries. The remaining sections of this study are as fol-
lows. Section 2 gives the theoretical underpinning of the global oil market; Sect. 3
provides the Bayesian granger causality model, while Sect. 4 discusses the empirical
results. Finally, Sect. 5 concludes.

2 OPEC and non‑OPEC countries in the global oil market

and theoretical issues

Price war is a situation where by two or more competitive firms reduce price of a com-
modity below that of their competitors in order to gain market share. The occurrence
of price war in the global oil market can result in imbalances and destabilization in the
global economy such that the welfare of oil producing countries are negatively affected.
This is because of the huge importance of oil supply and demand to the global econ-
omy (Kilian 2009). The avoidance of price war to ensure global economic stability is
the major reasons for the oil producing cartels to cooperate rather than compete. When
oil price is too low, oil importing countries gain while oil exporting countries gain
during high oil price. A reasonable global oil price is necessary to stabilize the global
economy. In March 2020, for instance, the two biggest oil exporting countries in the
world decided to pursue their individual interest rather than followed the OPEC + coop-
erative mechanism (Singh 2020). Their actions at that time created regional and global
economic imbalances coupled with the ravaging COVID-19 pandemic. There are sev-
eral countries that are players in the global oil market. Firstly, there are countries from
North America with United State as the highest oil producer. Secondly, we have coun-
tries from Common Wealth Independent States where Russia is the highest oil producer
in the region. There are also countries from Central America, Europe and Asia Pacific.
In the Middle East, Saudi Arabia is the highest oil producer in the region with average
daily production of 11 million barrels. Lastly, there are 12 oil exporting African coun-
tries with Nigeria as the largest oil exporter. All these regions together make up the
OPEC, OPEC + , non-OPEC, OECD, non-OECD and European Union (OPEC annual
Statistical Bulletin 2020). In terms of global production, the North America produces
25.9% of the global oil production with the US contributing 18.5%. The Middle East
produces 31.3% of the global production with Saudi Arabia contributing 12.2% (British

13
4608 Economic Change and Restructuring (2023) 56:4603–4628

Petroleum Statistical Review of world Energy 2022). The CIS region produces 15.4%
with Russia also contributing 12.2% of the share. On the other hand, Africa supplies
lesser production of 8.1%, Asia Pacific supplies 8.2%. Central America produces 6.6%
while Europe supplies the lowest with 3.8% of the global output (British Petroleum
Statistical Review of world Energy 2022). In term of global oil reserves, Saudi Arabia
holds the highest reserve of 17.2% of the global reserve, followed by Russia with 6.2%
and US with 4.0%. OPEC and non-OPEC have 70 and 30% share of global reserve.
The price war in the global oil market can be both traced to monopolistic and oli-
gopolistic structure of oil international market. Monopoly is a situation where a single
producer controls either the price or the output while oligopolistic market structure is
few producers that have the tendency of colluding to form a cartel. The presence of few
dominant oil producers like US, Saudi Arabia and Russia in the global oil market can
give room for either monopolistic or oligopolistic blocs. For instance, the monopolistic
tendency reflected in the shale oil supply by the US in 2016 for having a sole access
to the use of horizontal drilling and hydraulic fracturing technology. The increase in
the shale oil supply by the US brought down the global oil price from 112 USD per
barrel to 48 USD per barrel in that period. The fall in oil price was beneficial to the
US as a prior net oil importer. US access to fracking hydraulic technology has helped
her to increase the shale oil supply which makes her to have monopolistic power of
oil price and become a net oil exporter in the global oil market. On the other hand, the
oligopolistic structure could be seen in the light of cooperation between Saudi Arabia,
the leader of the OPEC and Russia, a member of OPEC + to keep their market share
even in the face of US increase in Shale oil. The agreement in oil supply between Rus-
sia and the OPEC countries has also led to a steady increase in oil price at 60 USD
per barrel between 2019 and 2020 (Ma et al. 2021). There have been controversies on
whether OPEC is a cartel or not. Cartel is a few competing firms that have agreed to
come together to maintain possible higher price through production control. Evidence
based studies like Khalid (2016) and Colgan (2014) described OPEC as a non-cartel
producer while Bremond et al. (2012) documented OPEC operating as a cartel. Smith
(2005), on the contrary concluded OPEC market structure to be in the middle of cartel
and non-cooperative Oligopoly.

3 The model

The BGN model begins with a vector autoregression (VAR) model structure specified
by Sims (1980) as:
Zt = B0 + B1 Zt−1 + B2 Zt−2 + ⋯ + B𝜌 Zt−𝜌 + 𝜇t (1)
where Zt and 𝜇t are random vectors with E(𝜇t ) = 0 and E(𝜇t 𝜇t� ) is a positive definite
matrix. Equation (1) is transformed by a lower triangular matrix A following Blan-
chard and Quah (1989) written as:
AZt = B0 + B1 Zt−1 + B2 Zt−2 + ⋯ + B𝜌 Zt−𝜌 + 𝜇t (2)
Inverting matrix A, Eq. (2) can be re-specified as:

13
Economic Change and Restructuring (2023) 56:4603–4628 4609

n
∑
Zt = A−1 B0 + A−1 Bi Zt−i + A−1 𝜇t (3)
i=1

Furthermore, Eq. (3) can be re-written in as:

n
∑
Zt = 𝛼0 + 𝛼i Zt−i + 𝜀t (4)
i=1

From standard point of view 𝛼0 = A−1 B0, 𝛼i = A−1 Bi and 𝜀t = A−1 𝜇t . Next, we
provide the representation of the acyclic graph structure.

3.1 Directed acyclic graph (DAG)

Directed acyclic graph is a graphical representation with nodes and directed edges
showing causal relations among several variables. There are four possible edge rela-
tions in DAG. The first is a non-directed edge (PQ) meaning that P is not related
with Q. The second is an undirected edge (P − Q) meaning uncertain causal direc-
tion between P and Q. The third is a direct edge (P → Q) meaning Q responds to
changes in Q. Where P is the independent variable and the parent (lead) and Q is
the dependent variable and the child (lag). The last is the directional edge (P ↔ Q)
meaning both A and B respond to each other simultaneously. Apart from the bivari-
ate model of the DAG, we can as well have a multivariate involving more than two
variables as (P → Q → R). This means that the relation between P and Q is con-
ditional on R. Where P and Q are ancestors to R while Q and R are descendants of
P . Following Ahelegbey et al. (2016), Ji et al. (2018), Yin and Ma (2018) and Ola-
yungbo (2019), the DAG network can be written in a VAR form in time dimension
as:
j
Zt−s → Zti with B∗ij ≠ 0 (5)
J is the past realization of Z which is equivalent to parent or ancestor and
where Zt−s
j
t
i is a future realization equivalent to child or descendant with s < t . Equation (5)
Zt
can be written in a more explicit chain form as:
j j j j
Zt−s →, … , Zt−3 → Zt−2 → Zt−1 → Zti (6)

Following the VAR specification in Eq. (1), the DAG representation can also be
specified as:
Zt = (G0 , Φ0 ) + (G1 , Φ1 )Zt−1 + (G2 , Φ2 )Zt−2 + ⋯ + (G𝜌 , Φ𝜌 )Zt−𝜌 + 𝜇t (7)
where (Gs , Φs ), with 0 ≤ s ≤ 𝜌 is the binary connectivity and the structural coef-
ficient of the lagged matrix. At any period of 0 ≤ s ≤ 𝜌, Gs,ij = 1 implies a causal
relation from Zt−s → Zti and Gs,ij = 0 implies no causal relation between Zt−s and Zti
j j

while Φs,ij implies the quantity of the causal effects of Zt−s

J
→ Zti .

13
4610 Economic Change and Restructuring (2023) 56:4603–4628

3.2 Bayesian inference

We begin the Bayesian inference with the standard posterior distribution given as:
f (Z|𝜃)f (𝜃)
f (𝜃|Z) = (8)
∫ f (Z|𝜃)f (𝜃)d𝜃

where f (Z|𝜃) is the likelihood function with X as the random data and 𝜃 as the
parameter to be estimated while f (𝜃) is the joint prior and the marginal likelihood
of the observed time series X is ∫ f (Z|𝜃)f (𝜃)d𝜃 . In proportionality form, Eq. (8)
becomes:
f (𝜃|Z) ∝ f (Z|𝜃)f (𝜃) (9)
Equation (9) implies posterior is likelihood multiplied by prior. The posterior dis-
tribution can further be written explicitly with the parameters of estimation as:

f (𝜇, Σ−1
Z
, G|Z) ∝ f (Z|𝜇, Σ−1
Z
, G)f (Σ−1
Z
|G) (10)

where 𝜇 is the mean, Σ−1 Z

is the variance and G is the graphical function. The mul-
tivariate normal distribution of the likelihood function with 𝜇 = 0 can be written as:
{ T
}
−nT T 1 ∑
(11)
−1 −1 −1 �
f (Z|ΣZ , G) = (2𝜋) 2 |ΣZ | 2 exp − (ZX ), Z t Zt
2 t=1

After defining the likelihood function, next is to derive the joint prior distribu-
tion, (Σ−1
Z
, G) expressed as f (G, Σ−1Z
) = f (G)f (Σ−1
Z
|G) with uniform distribution as
∑T
f (G)𝛼1 and the variance, ΣZ distributed as ΣZ W(𝜈, t=1 (Z0 Z0� )−1. With deriva-
−1 −1∼

tion of the joint prior distribution and the prior density, the posterior distribution can
then be stated as:
{ T
} { T
}
−nT T 1 −1 ∑ 1 (𝜈−n−1) 1 ∑
f (Σ−1
Z |Z) ∝ (2𝜋)
2 |Σ−1
Z | exp
2 − (Σ ), Z Z� , |Σ−1 | 2 exp − (Σ−1 ), Z Z�
2 Z t=1 t t Kn (𝜈, (Z0 Z0 )� Z 2 Z t=1 0 0
(12)
In Eq. (12) G is sampled using the multi Carlo Markov chain (MCMC) with the
Gibbs sampling to obtain the possible directed acyclic graph.

3.3 MCMC sampling

The MCMC sampling was initially proposed by Madigan and York (1995) and later
by Grzegorczyk and Husmeier (2008) and others. In the case of DAG, MCMC is a
process of generating a stationary distribution of a Markov chain on a space of
graphical structure. Its advantage over other sampling techniques is that it usually
converges to a representative of the true target posterior distribution. The procedure
involves the process of addition and deletion of a single edges in a Metropolis Hast-
ing. The MCMC and the MH algorithm involves sampling a new graph G∗ uniformly
conditioned on a graph G with acceptance probability given as: Q(G∗ |G) = min

13
Economic Change and Restructuring (2023) 56:4603–4628 4611

{ }
P(Z|G∗ ) P(G∗ ) Z(G|G∗ )
P(Z|G) P(G) Z(G∗ |G)
,1 where P(Z|G) is the likelihood function, P(G) is the prior
density and finally, Z(G∗ |G) is the proposal distribution. With convergence achieved
by the MCMC, the sample G∗ produces a graphical structure with which to make
inferences.

3.4 Bayesian Granger causality test

The Granger causality was propounded by Granger (1969) that past values of a
variable and other variable can impact significantly on itself. This causal process
is also extended to the graphical structure that the future realization (child or lag)
depends on the past realization (parent or lead) and other related past realization.
The Granger causality is a bivariate or pair wise Granger causality (PG-C) VAR 𝜌
structure stated as:

∑
𝜌
∑
𝜌−1
j
Zti = i
𝛼s Zt−s + 𝛽s Zt−s
s=1 s=0
(13)
j
∑
𝜌−1
∑
𝜌
j
i
Zt = 𝜋s Zt−s + 𝜃s Zt−s
s=0 s=1

Equation (13) implies instantaneous Z depends on its own past realization

i

and that of Z j at time t − 1 while instantaneous Z j also depends on its own past
realization and that of Z i at time t − 1. We test the following hypothesis whether
→ Zti , Zt−s → Zti , Zt−s → Zt and Zt−s → Zt for H0 ∶ 𝛼s = 𝛽s = 𝜋s = 𝜃s = 0 and
i j i j j j
Zt−s
H1 ∶ 𝛼s ≠ 𝛽s ≠ 𝜋s ≠ 𝜃s ≠ 0.

4 Empirical analysis

This section provides the empirical results starting from sources of the data, descrip-
tive analysis of the data, unit root tests, cointegration results, empirical results and
discussion of the results.

4.1 Data sources and variable description

In this study, we make use of annual global oil price and annual oil production of
three major world oil producing countries and OPEC oil production for the period
of 1965–2021. The oil price data used in this study are annual oil price quoted in
the international oil market, while the oil production data are annual oil production
of the sampled countries and OPEC. The oil price used is also annual oil price per
barrel quoted in US dollars in the global oil market. Specifically, the oil produc-
tion is measured as the average daily million barrels of oil produced and exported
to the international market yearly over the study period. The selected oil producing

13
4612 Economic Change and Restructuring (2023) 56:4603–4628

countries are Russia, US and Saudi Arabia. Both US and Russia are classified as
non-OPEC countries. The three countries, Russia, US and Saudi Arabia were cho-
sen being the first three highest oil producing countries in the world. The 13 OPEC
countries are also chosen because they produce almost 38 percent of the global oil
output (British Petroleum Statistical Review of World Energy 2022). In addition,
OPEC is a powerful controller of oil production and oil price in the world market.
The OPEC member countries are Algeria, Angola, Iran, Iraq, Kuwait, Libya, Nige-
ria, Saudi Arabia, United Arab Emirate, Qatar, Mexico, Gabon and Venezuela. The
choice of the starting date of 1965 is to capture the different periods of historical oil
supply disruptions in the analysis while the choice of the end date is due to the most
recent data availability. Both the oil price and production were sourced from the
British Petroleum statistical Review of world Energy (2022). The Brent and West
Texas Intermediate (WTI) oil price measured in US dollars were used in this study.
The brent oil price is chosen because it accounts for the two-third of oil pricing and
oil production in the global market while the WTI oil price is the pricing model
for the USA, a world leading supplier of oil in the international market. In conclu-
sion, Fig. 1 presented gives the visual changes in all the sampled variables over the
study period with Y axis showing the changes in quantity of the variables and X axis
showing the year.

4.2 Descriptive statistics

The descriptive statistics as presented in Table 1 show the data set in their unit val-
ues over the study period for 57 observations. The Table shows the average value of
34.43 USD and 33.73 USD for both the Brent and WTI oil price, respectively, over
the study period. A maximum price of 111.63 USD and 99.67 USD is shown for
Brent oil price and WTI oil price and a minimum price of 3 USD for both oil prices.
The large difference between the maximum and the minimum oil prices suggests
huge volatility in the oil price movements over the study period. The erratic move-
ment of the oil prices in Fig. 1 corroborates this assertion. Moreover, the positive
values of the skewness and the kurtosis values closer to 3 for both the oil prices indi-
cate the data distributions are a bit non-normally distributed. The 1% significance of
the Jarque Bera statistic confirms the non-normality of the oil prices’ distribution.
Considering the oil production data distribution, OPEC has the highest average oil
production 27,567,086 barrels, followed by US with almost 9,877,024 barrels, Rus-
sia with 9,361,053 barrels and Saudi Arabia with 8,321,067 barrels. The maximum
oil production for OPEC during the study period is 37,799,000 barrels, followed by
the US with 17,045,000 barrels, next is Saudi Arabia with 12,406,000 barrels and
lastly Russia with 12,403,000 barrels. The minimum barrels produced over the study
period are 13,709,000 barrels, 4,858,000 barrels 2,219,000 barrels and 6,783,000
barrels for OPEC, Russia, Saudi Arabia and US, respectively. These values sug-
gest that Saudi Arabia produced the minimum million barrels over the study period.
The negative skewness of the oil production for all the producers except USA sug-
gests positive distribution only for US oil production data during the study period.
However, the Jacque Bera statistic shows normal distribution only for OPEC oil

13
Economic Change and Restructuring (2023) 56:4603–4628 4613

Brent oil price WTI oil price

5.0 5.0

4.5 4.5

4.0 4.0

3.5 3.5

3.0 3.0

2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0
65 70 75 80 85 90 95 00 05 10 15 20 65 70 75 80 85 90 95 00 05 10 15 20

OPEC oil production Russia oil production

10.6 9.6

10.4 9.4

10.2 9.2

10.0 9.0

9.8 8.8

9.6 8.6

9.4 8.4
65 70 75 80 85 90 95 00 05 10 15 20 65 70 75 80 85 90 95 00 05 10 15 20

Saudi oil production US oil production

9.6 9.8

9.2 9.6

8.8 9.4

8.4 9.2

8.0 9.0

7.6 8.8
65 70 75 80 85 90 95 00 05 10 15 20 65 70 75 80 85 90 95 00 05 10 15 20

Fig. 1  Time series evolution of data used in the study

production and non-normal distribution for others. In summary, the descriptive sta-
tistics show that the data set of both the oil prices and oil production follows a ran-
dom and non-normal distribution over the study period.

13
4614 Economic Change and Restructuring (2023) 56:4603–4628

Table 1  Statistical distribution of the variables

Statistics Brent oil price WTI oil price OPEC Russia Saudi Arabia US

Mean 34.43 33.73 27,567.86 9361.53 8321.67 9877.24

Median 24.48 25 28,631 10,057 8983 9863
Maximum 111.63 99.67 37,799 12,403 12,406 17,045
Minimum 3 3 13,709 4858 2219 6783
Std. Dev 30.29 27.62 6829.2 2198.5 2844.43 2367.42
Skewness 1.14 1.00 −0.34 −0.47 −0.64 1.30
Kurtosis 3.4 3.02 2.04 1.85 2.30 4.83
Jarque Bera 12.79 9.60 3.26 5.23 5.02 24.00
Prob. 0.00*** 0.00*** 0.20 0.07* 0.08* 0.00***
Obs. 57 57 57 57 57 57

Note: *** and * signify significance at 1 and 10%

4.3 Stationarity property of the variables

In this study, we carry out preliminary tests of examining the stationarity property or
order of integration of the variables in order to avoid spurious results and to ensure
valid result of the parameter estimates. The Augmented Dickey Fuller (ADF 1979)
and Philip Perron (PP 1988) tests were done with 10 maximum lag length of the
variables in their log form and their results presented in Table 2. The results of both
the ADF and PP imply that all the series are stationary at first difference, mean-
ing they are all series of order 1 i.e., I(1) variables. To further ensure valid empiri-
cal results, we resort to structural break tests to capture the spikes observed in all
the variables used in this study (see Fig. 1). The reason for the spikes is not far
fetch from the oil supply disruptions and crises that have occurred from 1970s till
date. Also from econometric perspective, Bai-Perron (2003) affirmed that unit root
tests may fail to reject the null hypothesis in the presence of structural changes in
the underlying series. And from historical happenings, our data set must have wit-
nessed many structural breaks from many oil supply crises and shocks such as Arab

Table 2  Results of the unit root tests

Augmented Dickey Fuller (ADF) test Phillips–Perron (PP) test
Variables Levels First diff Status Variables Levels First diff Sta-
tus

Brent oil price −1.4993 −6.9062 I(1) Brent oil price −1.4555 −6.8542 I(1)
WTI oil price −2.6456 −7.6792 I(1) WTI oil price −2.626 −7.6954 I(1)
OPEC −2.1739 −5.6302 I(1) OPEC −2.1739 −5.7549 I(1)
Russia −2.6264 −3.3912 I(1) Russian −2.0045 −3.3599 I(1)
Saudi Arabia −2.2504 −6.2127 I(1) Saudi Arabia −2.5883 −6.2271 I(1)
US −1.6204 −4.3157 I(1) US 0.9912 −4.2337 1(1)

Note: The 5% critical value at first difference is −2.9155 for both the ADF and PP

13
Economic Change and Restructuring (2023) 56:4603–4628 4615

embargo of 1973, Iran–Iraq war of 1980 to 1988, 2011 US attack, 2016 shale oil
discovery, COVID-19 in 2020. Even the visual observation of the movement of the
data presented in Fig. 1 supports the presence of structural breaks. Therefore, as
a result of suspecting breaks in the data, we perform a test of multiply structural
breaks on each of the series to identify the break dates. The results are presented in
Table 3. The results are found to support the I(1) order of integration of all the series
with some years identified as break dates. The break dates from Table 3 conform to
the timing of the series of oil supply shocks. For instance, the break date of 2011 for
brent oil price coincides with the US September 2011 attack and WTI oil price also
corresponds to 2015 Russian–Ukraine conflict of Russia’s annexation of Crimea.
Lastly, 1973 Arab embargo spilled over to 1975 structural break date of OPEC pro-
duction, Iran–Iraq of 1980 to 1988 coincides with both Saudi Arabia and Russia’s
oil production break in 1982 and 1985, respectively. Finally, the US break date of
2016 correlates with Shale oil discovery in 2016.

4.4 Cointegration test

We go further to the cointegration test of Johansen and Juselius (1992) and

(Johansen 1988) after establishing that the variables are all I(1) even in the presence
of structural breaks. The multivariate unrestricted cointegration model can be speci-
fied as:

∑
𝜌−1
ΔZt = c0 + ΠZt−1 + Λi ΔZt−1 + 𝜀t (14)
i=1

In this study, Zt is a 5− vector of endogenous variables with Brent oil price,

OPEC, Russia, Saudi Arabia and US oil production on one hand and WTI oil price,
OPEC, Russia, Saudi Arabia and US oil production on another. Moreover, c0 is the
constant term, Π is the maximum number of independent vectors in the model while
𝜀t is the error term. If rk(Π) = 0, then there is no cointegration, however if rk(Π) = 5
then there is full rank cointegration otherwise if rk(Π) < 5 it implies short-ranked
cointegration. The long run matrix, Π = 𝛼𝛽 � where 𝛼 is matrix 5 × r matrix of
adjustment parameters while 𝛽 is 5 × r matrix of r cointegrating vectors. We present
the cointegration results in Table 4. The results for the test of cointegration among
Brent oil price, OPEC, Russia, Saudi Arabia and US imply one cointegrating vector

Table 3  Structural break result Variables Levels First diff Break date Sta-
tus

Brent oil price −3.6349 −7.8426 2011 I(1)

WTI oil price −3.7796 −8.9670 2015 I(1)
OPEC −4.1317 −6.2568 1975 I(1)
Russia −3.5224 −4.2390 1985 I(1)
Saudi Arabia −3.0502 −7.4215 1982 I(1)
US −3.1378 −8.5004 2016 I(1)

Note: The 5% critical value at first difference is −4.4436

13
4616 Economic Change and Restructuring (2023) 56:4603–4628

Table 4  Cointegration result (Brent oil price, OPEC, Russia, Saudi Arabia, US)
Cointegrating Eigen-value Trace statistics Critical value Prob.
vector

r≤0 0.4809 74.6336 69.8189 0.02**

r≤1 0.2747 38.5721 47.8561 0.28
r≤2 0.1962 20.9047 29.7971 0.36
r≤3 0.1224 8.9560 15.4947 0.37
r≤4 0.0318 1.7798 3.6415 0.18
Cointegrating Eigen-value Max-Eigen statistics Critical value Prob.
vector

r≤0 0.4809 36.0615 33.8769 0.03**

r≤1 0.2747 17.6674 27.5843 0.52
r≤2 0.1952 11.9467 21.1316 0.55
r≤3 0.1224 7.1782 14.2646 0.47
r≤4 0.0318 1.7798 3.8415 0.18

**Indicates significance at 5%

at r ≤ 0. The null hypothesis of no cointegration can be tested by comparing the

Eigen-value or the Trace statistics with the critical value. From Table 4, we can only
reject the null hypothesis of no cointegration at the cointegrating vector r ≤ 0 where
the Trace statistics’ value of 74.63 is greater than the critical value of 69.82 at 5%
significance level. One cointegrating vector can also be observed for the compari-
son between the Max-Eigen statistics and the critical value at 5% significance. We
also consider the cointegration test for WTI oil price, OPEC, Russia, Saudi Arabia
and US and found one cointegrating vector, r ≤ 0 at 5% significance in Table 5. In

Table 5  Cointegration result (WTI oil price, OPEC, Russia, Saudi Arabia, US)
Cointegrating Eigen-value Trace statistics Critical value Prob
vector

r≤0 0.4642 70.9421 69.8189 0.04**

r≤1 0.2685 36.6255 47.8561 0.37
r≤2 0.2003 19.4271 29.7971 0.46
r≤3 0.0985 7.1333 15.4947 0.56
r≤4 0.0257 1.4306 3.6415 0.23
Cointegrating Eigen-value Max-Eigen statistics Critical value Prob
vector

r≤0 0.4642 34.3157 33.8769 0.04**

r≤1 0.2685 17.1994 27.5843 0.56
r≤2 0.2003 12.2938 21.1316 0.52
r≤3 0.0985 5.7027 14.2646 0.65
r≤4 0.0257 1.4306 3.8415 0.23

**Indicates significance at 5%

13
Economic Change and Restructuring (2023) 56:4603–4628 4617

summary, the presence of one cointegrating vector indicates that at least one causal
direction exists among the variables of interest.

4.5 Bayesian Granger network (BGN) causality results

Having found the existence of cointegration among the variables, we proceed to the
Granger causality analysis using Bayesian techniques with the variables in logarithm
form. Our dynamic network analysis is different from other dynamic models like
the dynamic conditional correlation generalized autoregressive conditional het-
eroscedasticity (DCC-GARCH) model proposed by Engle (2002). While the DCC-
GARCH involves return and risk volatility, our Granger causal network is about pos-
terior causal inference among variable of interest. In addition, our dynamic BGN is
also different from network model by Wang et al. (2018), and recent ones by Sun
et al. (2022) and Wang et al. (2022). Specifically, Wang et al. (2018) paper uses fluc-
tuation conduction network (FCN) for Chinese stock markets, a transitional network
of oil price to Chinese price levels is employed by Sun et al. (2022) while Wang
et al. (2022) employ ExtremeRiskRank for spillover risk among 73 stock market in
six continents. In this case, we present both the multivariate instantaneous (MIN)
and the multivariate autoregressive (MAR) analysis. The MIN analysis is a contem-
poraneous network relation among the variables of interest in their current process
while the MAR analysis is a dynamic network relation among the variables in their
current and past realization. The lag length selection criterion, Schwarz criterion
(SC), shows lag one as the optimal lag length for the MAR network analysis. At esti-
mation, the variables are ordered as Zt = (US, Russia, Saudi, OPEC, Brent oil price)
and Zt = (US, Russia, Saudi, OPEC, WTI price).
The ordering of the variables can be done in any order as these arrangements do
not affect our results in any way. In the analysis, we run MCMC Gibbs algorithm
of 40,000 draws with each 20,000 draws for the MIN and MAR estimation. Using
the first 10,000 draws of each of the MIN and MAX analysis as burn-in, which rep-
resents 50% of each draw and following Geweke’s diagnostic test (Geweke 1992).
As regards the prior distribution, we follow the prior type of Minnesota and nor-
mal Wishart for Bayesian VAR estimation (Ahelegbey et al. 2016). The desired

Fig. 2  Multivariate instantaneous network structure with Brent oil price (MIN)

13
4618 Economic Change and Restructuring (2023) 56:4603–4628

Fig. 3  Multivariate autoregressive network structure with Brent oil price (MAR)

estimation convergence was achieved and the results displayed in Figs. 2 and 3. The
green (white) color indicates presence (absence) of causality running from the past
variable to the instantaneous one. It should be noted that the independent or explan-
atory variables are listed at the row side (X axis) while the dependent or response
variables are listed at the column side (Y axis) of the Figures. The MIN result from
Fig. 2 shows Saudi Arabia oil production to granger cause both OPEC oil produc-
tion and US oil production i.e., Saudit → OPECt and Saudit → USt . This implies oil
production from Saudi Arabia determines the barrels of oil produced both by OPEC
and US. First, this result may explain why higher oil price during the Gulf war in
1990 when Iraq invasion of Kuwait led to oil supply shortage. The supply cut made
Saudi Arabia through OPEC to fill the production gap and stabilize price by increas-
ing global oil supply. This may explain the reason why Saudi Arabia’s oil production
influences OPEC oil productions. As regards the significant relationship between
Saudi Arabia’s oil supply and the US oil supply. The Arab embargo in 1973 war is
when Saudi Arabia as an OPEC member ordered stoppage of oil trade to the US and
the western continent because of their support to Israel at that time may also explain
our findings. The effect of that oil trade embargo was immediate at that time. The
price of oil increased four times from 3 to 12USD per barrel, imposing skyrocketed
cost on US consumers. The role of energy-intensive industries declined in the USA
and the economy shrunk by 2.5% (Energy Information Administration 2023). US
congress had to pass subsidy laws to protect consumers from gasoline shortages and
high cost. The adverse effects of the supply cut from Saudi Arabia forced the US to
later establish the Strategic Petroleum Proven Reserve (SPPR), a large system of oil
tanks, to store oil for future contingencies (OPEC 2020). Given our line of discus-
sion, it is evident that the costs of price war, oil supply shocks and geopolitical crises
on oil producing countries and the global economy are usually huge and enormous.
This is much reason why coordinated oil supply actions are needed between OPEC
and non-OPEC countries to avoid any oil supply crisis and ensure oil global stability
for both oil producing and importing countries. In conclusion, the MIN result indi-
cates that Saudi Arabia is a leader, while US and OPEC are followers in the global

13
Economic Change and Restructuring (2023) 56:4603–4628 4619

oil market Our results support the findings of Alhajji and Huettner (2000) that Saudi
Arabia is a dominant player in the international oil market.
The MAR result in Fig. 3, on the other hand, shows the autoregressive response
of the current variables to the past variables. Unlike the MIN result where oil pro-
duction from Saudi Arabia determines US oil production, the MAR result, on the
other hand, shows past oil production from US to granger cause current oil produc-
tion from Saudi Arabia, Russia and US i.e., USt−1 → Saudit , USt−1 → Russiat and
USt−1 → USt . This result is in line with a recent study by Kisswani et al. (2022) that
investigated impact of non-OPEC oil supply on OPEC oil production and found long
run symmetric impact of non-OPEC production on OPEC production. The increase
in US oil supply to the global market from the discovery of shale oil in 2016 could
be the possible explanation for US oil supply to determine other global oil suppli-
ers like Russia and Saudi Arabia. The MAR result also shows past oil supply from
Russia to influence Brent oil price i.e., Russiat−1 → Brentt . The implication of this
outcome is that Russia can influence the movement of oil price in the global mar-
ket. Moreover, similar to the MIN result, past values of oil production from Saudi
Arabia also determines current oil production of OPEC, Russia and Saudi i.e.,
Saudit−1 → OPECt , Saudit−1 → Russiat and Saudit−1 → Saudit . And lastly, we find
both past values of Brent oil price and oil production from OPEC to influence their
current values i.e., OPECt−1 → OPECt and Brentt−1 → Brentt . In the MAR network
result, that is in the dynamic network we can infer that, USA is a leader to Saudi
Arabia, and Russia while Saudi Arabia emerges as a leader to both OPEC and Rus-
sia. Oil supply from Russia is found to determine Brent oil price. This makes Russia
to be a decisive factor in oil price determination in the global oil market. The impli-
cation of the overall result is that it will be more beneficial for the global oil market
and economy in term of economic stability for member countries and for the three
largest oil exporting countries, US, Saudi Arabia and Russia, to cooperate than to
compete.
We further present the MIN and the MAR result for US, Russia, Saudi, OPEC
and WTI oil price in Figs. 4 and 5. While the MIN result using WTI oil price as the
global oil price is the same with the MIN result of the Brent oil price. The MAR

Fig. 4  Multivariate instantaneous network structure with WTI oil price (MIN)

13
4620 Economic Change and Restructuring (2023) 56:4603–4628

Fig. 5  Multivariate autoregressive network structure with WTI oil price (MAR)

result with WTI oil price as the global oil price, on the other hand, is slightly dif-
ferent from the MAR result of the Brent oil price. In other words, the MAR result
with WTI shows the current US oil production responding to the past level of Rus-
sian oil production. This indicates that Russian oil production determines US oil
production. In addition, Russia’s current oil production responds to the past level of
both US oil production and Saudi oil production. Similarly, the results demonstrate
that Saudi Arabia’s current oil production responds to the past level of both US oil
production and its own past oil production level. Lastly, WTI oil price responds to
past value of Russia oil production. In summary, this result suggests that Russia and
US are interdependent in oil supply which means there should be mutual coopera-
tion between the two countries for mutual gain. The difference in the cost of oil
exploration between the two countries is enough to promote cooperation between
the two countries in order to maintain their market share. For instance, it requires
48–54 USD per barrel on the average to produce in the US, while with a lower price
of barrel below 30 USD, many oil wells will be closed in Russia. Only Saudi Arabia
has the production capacity to explore at much lower oil price of 10 USD per bar-
rel (Yang et al. (2022). Again, the result that Russia depends on Saudi Arabia’s oil
supply and Russia’s control of oil prices indicate that both countries should avoid
the repetition of the 2020 price war during COVID-19 when OPEC + agreement of
reducing oil supply could not be reached on time with devastating effects on oil pro-
ducing countries’ economies. The event of price war has deepened our understand-
ing that such move should be avoided completely by Saudi Arabia. Especially, now
that the reserves of non-OPEC countries have been built up.
In putting this study in the right perspective, it is observed that previous stud-
ies like Khalid (2016), Colgan (2014) and Heuttner (2002) have found OPEC as a
non-cartel organization from their studies of testing the cartel behavior of OPEC.
According to them, OPEC is a non-cartel organization due to lack of oil supply
coordination between OPEC and its member countries. However, on the contrary,
our study finds presence of supply coordination with Saudi Arabia predicting OPEC
oil production in both the instantaneous and autoregressive result. We can boldly
conclude from our findings that the market structure of OPEC is a cartel and oli-
gopoly. Furthermore, our results have also proved that Saudi Arabia as a coordinator

13
Economic Change and Restructuring (2023) 56:4603–4628 4621

of OPEC cannot achieve the objective of coordinating and unifying oil policies for
its member countries and global oil supply stability without cooperating with other
non-OPEC countries like US and Russia. And the cooperation of all the three major
oil producers is important given the adverse effects of non-cooperation in terms
national and global economic loss. Also, the claim by past studies such as Bremond
et al. (2012) and Yang et al. (2022) that OPEC determines global oil prices may
have been true in the past due to limited data usage. Our study has unveiled Rus-
sia as the major determinants of both brent and WTI oil prices. The increase in the
global oil prices above 100 USD after the invasion of Russia in Ukraine in February
2022 is a pointer to our claim.

4.6 Robustness analysis

We further carry out a robustness check on the variables of interest using the tra-
ditional VAR-based Granger causality model to see if different results will be
observed. The advantage of this Granger causality test over the BGN causality is
being a mediating Granger causality test. It is unlike the BGN causality test which is
pair wise causality test. The VAR representation with optimal lag length of one can
be specified as:
USt = 𝜔1 + 𝜔2 USt−1 + 𝜔3 Russiat−1 + 𝜔4 Saudit−1 + 𝜔5 OPECt−1 + 𝜔6 Brentoilprice + 𝜀1

Russiat = 𝛽1 + 𝛽2 USt−1 + 𝛽3 Russiat−1 + 𝛽4 Saudit−1 + 𝛽5 OPECt−1 + 𝛽6 Brentoilpricet−1 + 𝜀2

Saudit = 𝜂1 + 𝜂2 USt−1 + 𝜂3 Russiat−1 + 𝜂4 Saudit−1 + 𝜂5 OPECt−1 + 𝜂6 Brentoilpricet−1 + 𝜀3

OPECt = 𝜅1 + 𝜅2 USt−1 + 𝜅3 Russiat−1 + 𝜅4 Saudit−1 + 𝜅5 OPECt−1 + 𝜅6 Brentoilpricet−1 + 𝜀4

Brentoilpricet = 𝜃1 + 𝜃2 USt−1 + 𝜃3 Russiat−1 + 𝜃4 Saudit−1 + 𝜃5 OPECt−1 + 𝜃6 Brentoilpricet−1 + 𝜀5
(15)
where the variables are still defined as before. And the parameters are 𝜔i , 𝛽i , 𝜂i , 𝜅i , 𝜃i
and 𝜀1 , 𝜀2 , 𝜀3 , 𝜀4 , 𝜀5 are the error terms. The test of hypothesis of no Granger cau-
sality is that null hypothesis, H0 ∶ 𝜔i = 𝛽i = 𝜂i = 𝜅i = 𝜃i = 0. The MAR struc-
ture of the BGN causality model is similar to traditional Granger causality model
because of its autoregressive structure. In Table 6 using Brent oil price, we find
Russiat−1 → Brentt , Saudit−1 → OPECt and Saudit−1 → Russiat to be common cau-
sality between the traditional Granger causality and the BGN MAR result in Fig. 3.
This result implies that Russia’s oil production influences brent oil price while Saudi
Arabia’s oil predicts both OPEC oil production and Russia’s oil production. The tra-
ditional VAR result also confirms the BGN MAR result in Fig. 5 using WTI oil
price. The result shows Saudit−1 → OPECt , Saudit−1 → Russiat and Russia → WTI
as common causality between the traditional causality and the BGN MAR. The
result suggests that Saudi Arabia does not only determines OPEC oil production but
also Russia’s oil production. Russia’s oil production also determines WTI oil price.
Lastly, in verifying the validity and reliability of our results, we carry out a diag-
nostic test on the BGN and the traditional Granger causality model. We embark on

13
4622 Economic Change and Restructuring (2023) 56:4603–4628

in-sample forecast on the full sample period. In the estimation, the forecast value
and the actual value track each other to determine the forecast error. And from the
Appendix, Fig. 6 shows in-sample forecast error plot between the actual and the
forecast value. We present only the forecast error model of Brent oil price, OPEC,
Russia, Saudi Arabia, US. The model for WTI oil price, OPEC, Russia, Saudi Ara-
bia, US is not presented due to space constraint. It can be made available upon
request. It is obvious the gap between the graphs in Fig. 6 is small. The closeness
of the two graphs suggests lowest error in both our models and results. We can infer
from Fig. 6. that the Granger causality models employed in this study are adequate
and the results reliable for policy recommendations.

5 Conclusion and policy implications

Our study explores both the static and dynamic relationships between oil sup-
ply and oil price determination among OPEC and non-OPEC member countries
from the period of 1965 to 2021. The choice of the sample countries is borne
out of the leading roles the countries represent in the international oil market in
terms of their huge oil supply to the global oil market. After performing all the
preliminary tests, we employ both the BGN causality and the traditional VAR
causality in order to uncover the robust relationship between oil production and
oil price among the leading oil producers. This is an attempt to provide more
understanding into the global phenomena of oil price war, oil supply shocks,
geopolitical crises, GFCs and the COVID-19 pandemic. In order to further vali-
date our models and results, we carry out a model selection evaluation through
an in-sample forecast analysis. The closeness of both the actual plot and the
forecast plot proves the reliability of our models used and the results reported.
The static BGN model shows Saudi Arabia to determine both OPEC and US
oil production. Furthermore, both the dynamic BGN and the traditional Granger
causality model reveal Saudi Arabia’s oil production to determine both OPEC
and Russia’s oil production. US oil supply predicts Saudi Arabia’s oil produc-
tion, while Russia and the US oil supply are interdependent. Interestingly, only
Russia’s oil production is found to determine both the brent and WTI oil prices
(Table 7).
The following implications emerge from our study. First, a form of market
collusion and coordination seem to exist between Saudi Arabia and the OPEC
member countries given the dependence of OPEC oil output on Saudi Arabia’s
oil supply. This is understandable from the periodic of meetings held to dictate
oil production quotas to OPEC member countries. Secondly, Saudi Arabia is a
leader to both Russia and the US in the oil international market as shown by the
dependence of both Russia and US oil production on Saudi Arabia’s oil supply in
the dynamic and static model. By implication, both Russia and US are followers
to Saudi Arabia in the global oil market. Our study further finds structural rela-
tionship between Russia and US with US oil production predicting Russia’s oil
production in the international oil market and vice versa. By implications, Rus-
sia and the US are interdependence in oil supply and the expectation is that the

13
 Table 6  VAR Granger causality Wald test for (US, Russia, Saudi, OPEC, Brent oil price)
Dep. Var US Russia Saudi OPEC Brent
Indep. Var Chi-square p-val Chi- square p-val Chi-square p-val Chi-square p-val Chi-square p-val

US − − 1.46 0.23 0.33 0.58 0.77 0.37 5.72** 0.02

Economic Change and Restructuring (2023) 56:4603–4628

Russia 1.99 0.16 − − 0.68 0.41 0.37 0.55 6.50** 0.01

Saudi 12.48*** 0.00 9.22*** 0.00 − − 12.53*** 0.00 0.76 0.38
OPEC 7.77*** 0.00 14.21*** 0.00 17.70*** 0.00 − − 1.46 0.22
Brent 18.66*** 0.00 1.46 0.29 0.29 0.59 0.05 0.82 − −

Note: *** and ** denote 1 and 5% significance level, respectively

4623

13
4624 Economic Change and Restructuring (2023) 56:4603–4628

two countries should cooperate for mutual growth and benefit. In addition, the
three leading oil countries should avoid cold war and price war of using politics
and economic powers to cause oil supply disruptions and oil price volatility in
the international market. Also, the market share of each country and each group
should be strictly maintained to avoid oil glut that can drive down price to unrea-
sonable level. The US has the highest global market share while Saudi Arabia
owns the largest share of the global reserve. Saudi Arabia as the holder of the
largest global reserve should not decide to increase her market share to disrupt
the market because of high reserve capacity or pursuit of profits. It is advisable
that each country should respect individual country’s position, capacity and sta-
tus in the market for global stability. Lastly, Russia’s oil supply determines the oil
prices of both Brent and WTI in the global market. We can recall that oil prices
increased above US$100 immediately Russia invaded Ukraine in February 24th,
2022 precisely. This is a clear indication that Russia is a global determinant of
oil prices. Although, this study could not capture the ongoing Russia–Ukraine
war due to unavailability of data up to the present period for other variables apart
from the global oil prices. It is hoped that extending the time scope to the pre-
sent period is a direction for future studies. Further studies can also be done by
extending the sample countries to more major non-OPEC countries like Kazakh-
stan, Malaysia and Oman. Other major oil producing countries in North America,
Asia and Europe can also be included in future studies to capture more dynamics
in the global oil market. Moreover, other studies could further explore the com-
plex and dynamic relationship between oil supply and prices, including the role
of various factors and the impact of OPEC’s policies, alternative energy sources,
economic development, and recent geopolitical war.
This study therefore concludes that though Saudi Arabia is a dominant and
leading oil player in global market, yet Russia’s oil production matters in the
determination of the global oil prices while the US oil supply acts as a global
stabilizer. Based on the findings and conclusion of our study, the following pol-
icy recommendations are made. Firstly, it highlights the importance of coordina-
tion and cooperation among these three major oil-producing countries to ensure
stability and sustainability in the global oil market. The three-leading oil-pro-
ducing countries must work together to avoid supply gluts and market disrup-
tions that could lead to sudden price fluctuations. Secondly, it emphasizes the
importance of monitoring and regulating the global oil market to ensure fair
competition, prevention of price manipulation, and safeguard the interests of
consumers. Finally, the finding that the US oil supply acts as a global stabilizer
implies that the US has a significant role to play in maintaining stability in the
global oil market. This could involve measures such as strategic oil reserves,
flexible production policies, and diplomatic efforts to maintain stable relation-
ships with other major oil-producing countries. Overall, the finding that Saudi
Arabia, Russia, and the USA each has a unique role in the global oil market
highlights the need for a coordinated and multifaceted approach to ensure stabil-
ity and sustainability in the global energy sector.

13
Economic Change and Restructuring (2023) 56:4603–4628 4625

Appendix

See Fig. 6.

Brent oilprice OPEC

3.6 10.4
3.2 10.2
2.8
2.4 10.0

2.0 9.8
1.6
9.6
1.2
0.8 9.4
65 70 75 80 85 90 95 00 05 10 15 20 65 70 75 80 85 90 95 00 05 10 15 20
Actual Actual LOILPROD_OPEC (Baseline)
LOIL_PRICE_BRENT (Baseline)

Russia Saudi Arabia

9.4 9.2

9.2 8.8
9.0
8.4
8.8
8.0
8.6

8.4 7.6
65 70 75 80 85 90 95 00 05 10 15 20 65 70 75 80 85 90 95 00 05 10 15 20
Actual LOILPROD_RUS (Baseline) Actual LOILPROD_SAUDI (Baseline)

US
9.6
9.5
9.4
9.3
9.2
9.1
9.0
65 70 75 80 85 90 95 00 05 10 15 20
Actual LOILPROD_US (Baseline)

Fig. 6  In-sample forecast error for (US, Russia, Saudi, OPEC, Brent oil price) model

13
 4626

13
Table 7  VAR Granger causality Wald test for (US, Russia, Saudi, OPEC, WTI oil price)
Dep. Var US Russia Saudi OPEC WTI
Indep. Var Chi-square p-val Chi-square p-val Chi-square p-val Chi-square p-val Chi-square p-val

US − − 0.51 0.47 0.27 0.60 0.51 0.48 5.99** 0.01

Russia 1.69 0.19 − − 0.54 0.46 0.63 0.4 6.91** 0.01
Saudi 11.82*** 0.00 8.44*** 0.00 − − 13.17*** 0.00 1.15 0.28
OPEC 7.65** 0.01 14.08*** 0.00 17.61*** 0.00 − − 1.69 0.19
WTI 14.99*** 0.00 1.56 0.21 0.14 0.71 0.31 0.58 − −

Note: *** and ** denote 1 and 5% significance level, respectively

Economic Change and Restructuring (2023) 56:4603–4628
Economic Change and Restructuring (2023) 56:4603–4628 4627

Acknowledgements We thank the two anonymous reviewers for their contributions. The usual disclaimer
applies, any error is from the authors.

References
Ahelegbey DF, Billio M, Casarin R (2016) Bayesian graphical models for structural vector autoregressive
processes. J Appl Economet 31:357–386
Alhajji A, Huettner D (2000) OPEC and other commodity cartels: a comparison. Energy Policy
28(15):1151–1164
Bai J, Perron P (2003) Computation and analysis of multiple structural change models. J Appl Economet
18:1–22
Blanchard O, Quah D (1989) The dynamic effects of aggregate demand and supply disturbances. Am
Econ Rev 79:655–673
British Petroleum Statistical Review of World Energy (2022). http://​www.​bp.​com/​stati​stica​lrevi​ew,
Accessed on 20th August, 2022.
Branger N, Flacke RN, Graber N (2019) Monopoly power in the oil market and the macroeconomy.
Energy Econ. https://​doi.​org/​10.​1016/j.​eneco.​2019.​104597
Bremond V, Hache E, Mignon V (2012) Does OPEC still exist as a cartel? An empirical investigation.
Energy Econ 34(1):125–131
Cai Y, Zhang D, Chang T, Lee C (2022) Macroeconomic outcomes of OPEC and non-OPEC oil supply
shocks in the Euro area. Energy Econ. https://​doi.​org/​10.​1016/j.​eneco.​2022.​105975
Colgan J (2014) The emperor has no clothes: the limits of OPEC in the global oil market. Int. Organ
68(3):599–632
Consumer News and Business Channel (CNBC) (2020) Accessed 29th April, 2023, https://​www.​cnbc.​
com/​2020/​04/​23/​how-​the-​oil-​price-​capit​ulati​on-​will-​hit-​niger​ia-​saudi-​arabia-​and-​other-​major-​expor​
ters.​html.
Dickey DA, Fuller WA (1979) Distribution of the estimators for autoregressive time series with a unit
root. J Am Stat Assoc 74:427–431
Dreger D, Holodilin KA, Ulbricht D, Fidrmuc J (2015) Between the hammer and the anvil: the impact
of economic sanctions and oil prices on Russia’s Ruble. J Comp Econ. https://​doi.​org/​10.​1016/j.​jce.​
2015.​12.​010
Engle R (2002) Dynamic conditional correlation. J Bus Econ Stat 20(3):339–250. https://​doi.​org/​10.​
1198/​07350​01022​88618​487
Geweke J (1992) Evaluating the accuracy of sampling-based approaches to the calculation of posterior
moments. In: Bernardo JM, Berger JO, David AP, Smith AFM (eds), Bayesian Statistics 4. Oxford
University Press, Oxford, pp. 169–193.
Granger CWJ (1969) Investigating causal relations by econometric models and cross-spectral methods,
Econometrica, 357–84.
Grzegorczyk M, Husmeier D (2008) Improving the structure MCMC sampler for Bayesian networks by
introducing a new edge reversal move. J Mach Learn 71:265–305
Khalid Mk (2016) Does OPEC act as a cartel? Empirical investigation of coordination behaviour. Energy
Policy. https://​doi.​org/​10.​1016/j.​enpol.​2016.​07.​013
Kilian L (2009). Not all oil shocks are alike: disentangling demand and supply shocks in the crude oil
market. Am Econ Rev, 1053–1069.
Kissiwani KM, Lahiani A, Mefteh-Wali S (2022) An analysis of OPEC oil production reaction to Non-
OPEC oil supply. Resources Policy. https://​doi.​org/​10.​1016/j.​resou​Pol.​2022.​102653
Ji Q, Bouri E, Gupta, R, Roubaud D (2018) Network causality structures among Bitcoin and other finan-
cial assets: a directed acyclic graph approach. Quart Rev Econ 70:203–213. https://​doi.​org/​10.​
1016/j.​qret.​2018.​05.​016
Johansen S (1988) Statistical analysis of cointegration vectors. J Econ Dyn Control 12(2–3):231–254.
https://​doi.​org/​10.​1016/​0165-​1889(88)​90041-3
Johansen S, Juselius K (1992) Testing structural hypotheses in a multivariate cointegration analysis of the
PPP and the UIP for UK. J Econ 53(1–3):211–244. https://​doi.​org/​10.​1016/​0304-​4076(92)​90086-7
Madigan, York J (1995) Bayesian graphical models for discrete data. Int Stat Rev 63:215–232

13
4628 Economic Change and Restructuring (2023) 56:4603–4628

Ma RR, Xiong T, Bao Y (2021) The Russia-Saudi Arabia oil price war during the COVID-19 pandemic.
Energy Econ. https://​doi.​org/​10.​1016/j.​eneco.​2021.​105517
Olayungbo DO (2019) Bayesian graphical model application for Monetary policy and macroeconomic
performance in Nigeria. Adv Novel Appl. https://​doi.​org/​10.​5772/​Intec​hopen​87994
Organisation Petroleum Exporting Countries (OPEC) (2012) Annual Statistical Bulletin, Vienna, Austria.
http://​www.​opec.​org.
Organisation Petroleum Exporting Countries (OPEC) (2020) Annual Statistical Bulletin, Vienna, Austria.
Retrieved from http://www.opec.org/opec-web.
Phillips PCB, Perron P (1988) Testing for a unit root in time series regressions. Biometrica 75:335–346
Sims CA (1980) Macroeconomics and reality. Econometrica 48:1–48
Singh AK (2020) The oil war of 2020 between Saudi Arabia and Russia: examining the underlying geo-
political and geoeconomic compulsions. Indian J Asian Affiars 33(1):24–42
Smith J (2005) Inscrutable OPEC? Behavioural tests of the cartel hypothesis. Energy J 26(1):51–82
Statista (2022) Monthly West Texas Intermediate and Brent oil price. Available online at: https://​www.​
stati​sta.​com/​stati​stics/​466293/​lowest-​crude-​oil-​prices-​due-​to-​covid-​19/. Accessed July 26th, 2022.
Sun Q, Wang Z, Jia N (2022) Revisiting the dynamic response of chinese price level to crude oil price
shocks based on a network analysis method. Entropy 24:944. https://​doi.​org/​10.​3390/​e2407​0944
This day newspaper in Nigeria (2020) Accessed 9th September, 2023. https://​www.​thisd​aylive.​com/​
index.​php/​2019/​10/​03/​house-​appro​ves-​57-​barrel-​as-​crude-​oil-​bench​mark-​for-​2020?​utm_​source=​
veooz​&​utm_​medium=​refer​ral
Umechukwu C, Olayungbo DO (2022) US oil supply shocks and economies of oil exporting African
countries; A GVAR-oil resource analysis. Resources Policy 75. https://​doi.​org/​10.​1016/j.​resou​rpol.​
2021.​102480.
US Energy Information Administration (2023). https://​www.​eia.​gov.
US-Saudi Business Council (2020) 8081 Wolftrap, Road, Suite 300 Vienna, VA 22182 and Riyadh,
Olaya Towers, Tower B-18th floor, Kingdom of Saudi Arabia, https://​ussau​di.​org/​saudi-​arabi​as-​
2020-​budget-​review/. Accessed 29th April, 2023.
Wang Z, Gao X, Tang R, Liu X, Sun Q, Chen Z (2018) Identifying influential nodes based on fluctuation
conduction network model. Physica A. https://​doi.​org/​10.​1016/j/​physa.​2018.​09.​078
Wang Z, Gao X, Huang S, Sun Q, Chen Z,Tang R, Di Z (2022) Measuring systemic risk contribution
of global stock markets: A dynamic tail risk network approach. https://​doi.​org/​10.​1016/j.​irfa.​2022.​
102361.
Yang TY, Chen SH, Yang YT, Chang HH (2022) Relationship between crude oil price and production
levels-an empirical study of OPEC and non-OPEC states. Front Environ Sci. https://​doi.​org/​10.​
3389/​fenvs.​2022.​933431
Yin L, Ma X (2018) Causality between oil shocks and exchange rate: a Bayesian, graph-based VAR
approach. Physica A. https://​doi.​org/​10.​1016/j.​physa.​2018.​05.​064

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and
applicable law.

13
Economic Change & Restructuring is a copyright of Springer, 2023. All Rights Reserved.