Accepted Manuscript

The Relationship between Energy Consumption, Economic Growth, and CO2

Emission in MENA Countries: Causality Analysis in the Frequency Domain

Muhammed Sehid Gorus, Mucahit Aydin

PII: S0360-5442(18)32350-8

DOI: 10.1016/j.energy.2018.11.139

Reference: EGY 14237

To appear in: Energy

Received Date: 22 February 2018

Accepted Date: 28 November 2018

Please cite this article as: Muhammed Sehid Gorus, Mucahit Aydin, The Relationship between
Energy Consumption, Economic Growth, and CO2 Emission in MENA Countries: Causality
Analysis in the Frequency Domain, Energy (2018), doi: 10.1016/j.energy.2018.11.139

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

The Relationship between Energy Consumption, Economic Growth, and CO2 Emission
in MENA Countries: Causality Analysis in the Frequency Domain

[Muhammed Sehid Gorus] 1

[Mucahit Aydin] 2,*

1[Faculty of Political Sciences, Department of Economics, Ankara Yıldırım Beyazıt

University, Esenboğa Campus, Çubuk/Ankara, Turkey]

2[Faculty of Political Sciences, Department of Econometrics, Sakarya University, Esentepe

Campus, Serdivan/Sakarya, Turkey]

Declarations of interest: none.

*corresponding author
 ACCEPTED MANUSCRIPT

The Relationship between Energy Consumption, Economic Growth, and CO2 Emission
in MENA Countries: Causality Analysis in the Frequency Domain

Abstract

The study investigates the causal relationship between energy consumption, economic growth,
and CO2 emission through both single- and multi-country Granger causality analysis in the
frequency domain considering eight oil-rich MENA countries; namely, Algeria, Egypt, Iran,
Iraq, Oman, Saudi Arabia, Tunisia, and the United Arab Emirates for the period 1975-2014.
The panel frequency domain analysis detects more causal nexus between variables across
different frequencies compared to the time domain causality. According to panel causality test
results, energy conservation policies do not have an adverse effect on economic growth both in
the short- and intermediate-run while their effects are negative in the long-run. Moreover,
policies to control air pollution can be designed by policymakers because of the absence of the
causal nexus between economic growth and CO2 emission.

Keywords: CO2 emission, economic growth, energy consumption, frequency domain approach

1. Introduction

The environmental economics and energy economics have come into prominence for a few
decades due to harsh environmental issues and global warming in the world. The relationship
between energy consumption, GDP (Gross Domestic Product) per capita, and carbon dioxide
(CO2) emission has been subject to many academic studies. However, empirical literature has
focused on developed countries (i.e., European countries, OECD countries) while there is a
little attention on the MENA (the Middle East and North African) region. World Bank (2008)
report states that MENA countries are on the unsustainable development path and they spend
their revenues which getting from crude oil export on oil extraction instead of investing these
revenues to provide long-run economic growth. The report also states that CO2 emission
increases and per capita water resources decrease in the region. Thus, the empirical studies take
into consideration energy and environmental issues in MENA countries in recent years.

1
 ACCEPTED MANUSCRIPT

There are several research strands in the literature about the energy and the environmental
economics. The first strand is Environmental Kuznets Curve (EKC) hypothesis which
investigates the impact of income per capita on the environmental degradation in the long-run.
This hypothesis asserts that environmental pollution increases with GDP per capita in the early
stages of economic growth, and then it decreases after a threshold level of income with
composition and technique effects (Grossman and Krueger, 1991; Acaravci and Ozturk, 2010).
Secondly, the Pollution Haven Hypothesis (PHH) examines the relationship between Foreign
Direct Investment (FDI) inflow and pollution level. It asserts that trade and investment
liberalization policies increase environmental degradation in developing countries since the
pollution-intensive production shifts from developed countries that have stringent
environmental policies to emerging countries that have relatively weak environmental
regulations (Copeland, 2010). Thirdly, unit root analysis is mostly used to determine
stationarity properties of the series (i.e., GDP per capita, energy consumption, CO2 emission).
Unit root analysis provides beneficial information to the policymakers since it shows the effects
of shocks (permanent or temporary) on the series.

The last strand investigates the relationship between energy consumption (i.e., total energy,
crude oil, coal, natural gas, electricity, renewable energy) and income level. The nexus between
aforementioned variables can be examined both cointegration and causality analysis. There are
four main hypotheses about the causative connection between energy consumption and
economic growth in the energy economics literature; namely, conservation hypothesis, growth
hypothesis, neutrality hypothesis, and feedback hypothesis (Ozturk, 2010; Ahmed and Azam,
2016; Hasanov et al., 2017):

i. Conservation Hypothesis assumes that a uni-directional causality runs from

economic growth to energy consumption (Y→EC). It asserts that economic growth
is the main factor of energy consumption and increase in income level causes
increase in energy consumption. This hypothesis emphasizes that energy
conservation policies do not have an adverse effect on economic growth. A county
can be called as less energy-dependent when conservation hypothesis is valid.
ii. Growth Hypothesis assumes that a uni-directional causality runs from energy
consumption to economic growth (EC→Y). Energy is an important input in the
production process according to this hypothesis. Thus, energy conservation policies
have negative effects on economic growth. A county can be called as energy-
dependent when growth hypothesis is valid.

2
 ACCEPTED MANUSCRIPT

iii. Feedback Hypothesis states that there is a bi-directional causal relationship between
energy consumption and economic growth (EC↔Y). So, these variables are jointly
determined and affected at the same time.
iv. Neutrality Hypothesis asserts that there is no causal link between energy
consumption and economic growth. So, conservative or expansive energy policies
do not affect economic growth. Thus, energy conservation policies can be
implemented easily when this hypothesis is valid in a country.

In this study, we examine the causal relationship between energy consumption, economic
growth, and CO2 emission in the short-, the intermediate-, and the long-run for eight oil-rich
MENA countries. The study uses both single- and multi-country Granger causality analysis in
the frequency domain from 1975 to 2014. Our sample consists of Algeria, Egypt, Iran, Iraq,
Oman, Saudi Arabia, Tunisia, and the United Arab Emirates (UAE). We choose oil-rich MENA
countries because of their abundant natural resource reserves and production volumes.
According to British Petroleum (2017) report, the Middle Eastern countries and the African
countries produce 43.06% of the crude oil in the world in 2016. Our sample is chosen with
regard to the data availability, and we select only North African countries rather than
considering whole African countries. After these adjustments, crude oil production volume of
these eight MENA countries become 31.28% of the world.

The main contributions of this study to the literature are three-fold. Firstly, to our best
knowledge, there is not any panel Granger causality analysis in the frequency domain that
provides the short-, the intermediate-, and the long-run causality results considering specifically
for eight oil-rich MENA countries in the literature. This recently developed frequency domain
test for panel data analysis computes the Granger causality for each individual frequency
component in place of calculating a single Granger causality measure for the entire causal
association (Croux and Reusens, 2013). Secondly, we conduct both single- and multi-country
Granger causality analysis. So, policymakers can implement country-based and/or region-based
economic policies. Thirdly, this study considers not only applicability of energy conservation
policies but also emission conservation policies in these countries.

The outline of this article as follows: Section 2 presents the empirical literature while Section
3 provides data and methodology. Empirical findings of the paper presented in Section 4. In the
last section, we conclude the study and provide some policy suggestions for policymakers.

2. Literature Review

3
 ACCEPTED MANUSCRIPT

The energy-growth nexus is extensively studied in the literature. The empirical literature shows
that the studies about the causal relationship between energy consumption and economic growth
give mixed results due to differences in the econometric methodologies conducted, type of
energy consumption used, and time span and countries/region chosen. Moreover, in some
studies, the causal relationship between CO2 emission, energy consumption, and economic
growth are investigated (i.e., Soytas et al., 2007; Acaravci and Ozturk, 2010; Al-Mulali, 2011;
Omri, 2013; Ozcan, 2013; Farhani and Shahbaz, 2014; Salahuddin and Gow, 2014; Alshehry
and Belloumi, 2015). The review of the literature is presented in following subsections. Besides,
the sample, time span, methodology, variables, and the main results of the selected studies are
displayed in Table 1.

2.1. Preliminary Studies

Kraft and Kraft (1978) examine the causal link between energy consumption and economic
growth for the United States for the time period 1947-1974. Empirical results indicate that the
conservation hypothesis is valid for the U.S. economy. Besides, Akarca and Long (1980)’s and
Yu and Hwang (1984)’s findings support the neutrality hypothesis in the U.S. economy over
the period of 1950-1970 and of 1947-1979, respectively. Yu and Choi (1985) indicate that the
neutrality hypothesis is valid for the U.K., the U.S., and Poland. Besides, the growth and the
conservation hypotheses are valid for Philippines and Korea from 1950 to 1976, respectively.
In addition, Erol and Yu (1987) provide mixed results for six advanced countries; namely,
Japan, Italy, Germany, Canada, France, and the U.K. over the period between 1952 and 1982.

Nachane et al. (1988) examine the energy-GDP nexus in 16 selected (11 developing and five
developed) countries from 1950-1951 to 1984-1985. Empirical findings provide mixed results
for this set of countries. Moreover, Hwang and Gum (1991) find that there is a bi-directional
Granger causality linkage between energy consumption and economic growth in Taiwan for
the period 1961-1990.

2.2. Recent Studies

Mehrara (2007) examines the causal linkage between energy consumption and economic
growth in 11 oil exporting countries from 1971 to 2002. Empirical results of the study suggest
that there is a one-way strong causality running from economic growth to energy consumption.
Moreover, Soytas et al. (2007) find that the neutrality hypothesis is valid in the U.S. economy
for the period 1960-2004. So, there is no causal relationship between energy consumption and
economic growth, however, there is a one-way Granger causality from energy use to CO2

4
 ACCEPTED MANUSCRIPT

emission. Acaravci and Ozturk (2010) investigate the causal relationship between energy
consumption, economic growth, and CO2 emission in 18 European countries. Empirical
findings provide mixed results. The overall results show that conservative policies related to
energy consumption and CO2 emission do not affect economic growth negatively. Bildirici and
Kayıkçı (2013) find that there is a bi-directional causal relationship between oil production and
economic growth both in the short- and long-run in four Eurasian countries during the time
period of 1993-2010.

Bozoklu and Yilanci (2013) use Granger causality analysis in the frequency domain to
determine the direction of causation for 20 OECD countries. Empirical findings support that
there are mixed results regarding the direction of causation between energy consumption and
economic growth. Alshehry and Belloumi (2015) find one-way causality from energy
consumption to economic growth in Saudi Arabia for the period 1971-2010. Also, Ahmed and
Azam (2016) examine the causal relationship between energy consumption and economic
growth for 119 countries through Granger causality analysis in the frequency domain. The
feedback hypothesis is valid for only 18 countries while the growth hypothesis is valid for 25
countries. Besides, 40 countries confirm the existence of conservation hypothesis, and 36
countries suggest neutrality hypothesis. Mutascu (2016) employs bootstrap panel Granger
causality method to examine the causal link between energy consumption and economic
growth. Empirical findings provide mixed results for G7 countries during 1970-2012.

Tang et al. (2016) find that the growth hypothesis is valid for Vietnamese economy for the
period 1971-2011 according to the Toda-Yamamoto procedure. On the other hand, Destek and
Aslan (2017) get mixed causality results for 17 developing countries. The study examines the
linkage between renewable and non-renewable energy consumption and economic growth
through bootstrap panel Granger causality methodology. Hasanov et al. (2017) find that there
is a uni-directional causality running from primary energy consumption to economic growth
both in the short-run and in the long-run in 10 Eurasian countries during the time period of
1997-2014. This finding supports the growth hypothesis for this set of countries. Lastly, Nyasha
et al. (2018) find that the conservation hypothesis is valid for Ethiopia over the period 1971-
2013.

2.3. Studies about MENA Countries

Al-Iriani (2006) finds that there is a uni-directional causality from GDP to energy consumption
in six Gulf Cooperation Council (GCC) countries over the period of 1971-2002. Al-Mulali

5
 ACCEPTED MANUSCRIPT

(2011) find that there is a bi-directional causal nexus between energy consumption, economic
growth, and CO2 emission in both the short- and the long-run in 17 MENA countries for the
period of 1980-2009. Omri (2013) finds that there is a two-way causal nexus between energy
consumption and economic growth in 14 MENA countries over the period of 1990-2011. This
finding supports the feedback hypothesis in the region. Besides, there is also a uni-directional
causality running from energy consumption to CO2 emission and bi-directional causality
between economic growth and CO2 emission. Furthermore, Ozcan (2013) demonstrates that
there is a uni-directional Granger causality from economic growth to energy consumption in 12
MENA countries over the period between 1990 and 2008. Thus, the energy conservation
policies are applicable in these countries. Besides, these 12 MENA countries can be called as
less energy-dependent.

Farhani and Shahbaz (2014) investigate the causal nexus between CO2 emission, electricity
(renewable and non-renewable) consumption, and economic growth in 10 MENA countries for
the period between 1980 and 2009. According to empirical findings, there are uni-directional
causalities from renewable and non-renewable electricity consumption to CO2 emission in the
short-run. Besides, a uni-directional causal relationship exists from economic growth to CO2
emission. In the long-run, a bi-directional relationship is confirmed between electricity
(renewable and non-renewable) consumption and CO2 emission. Besides, Salahuddin and Gow
(2014) find that the conservation hypothesis is valid for six GCC countries for the period of
1980-2012. In other words, the one-way Granger causality exists from economic growth to
energy consumption. Lastly, Kahia et al. (2017) examine the causal relationship between
economic growth, renewable energy use, non- renewable energy use, real gross fixed capital
formation, and labor force for 11 MENA oil importing countries for the period between 1980
and 2012. Empirical results demonstrate that the feedback hypothesis is valid for both
renewable energy and non- renewable energy.

Table 1. Overview of Studies on the Energy Consumption and Economic Growth

Country(s),
Author(s) Methodology(s) Variables Main Results
Time period

A: Preliminary Studies
Energy
Kraft and Kraft Unites States,
Granger causality consumption, Y→EC
(1978) 1947-1974
economic growth
Energy
Akarca and Long United States,
Sims’ technique consumption, Y―EC
(1980) 1950-1974
economic growth

6
 ACCEPTED MANUSCRIPT

Energy
Yu and Hwang United States,
Sims’ technique consumption, Y―EC
(1984) 1947-1979
economic growth
5 selected Energy
Yu and Choi Sims’ technique,
countries, consumption, Mixed results
(1985) Granger causality
1950-1976 economic growth
6 advanced Energy
Sims’ technique,
Erol and Yu (1987) countries, consumption, Mixed results
Granger causality
1952-1982 economic growth
16 selected Cointegration, Energy
Nachane et al.
countries, 1950- Sims’ techique and consumption, Mixed results
(1988)
1985 Granger causality economic growth
Cointegration, error Energy
Hwang and Gum Taiwan,
correction model, consumption, Y↔EC
(1991) 1961-1990
Granger causality economic growth

B: Recent Studies
11 oil exporting Energy
Mehrara (2007) countries, Panel cointegration consumption, Y→EC
1971-2002 economic growth
Energy
United States, Toda-Yamamoto consumption, Y―EC,
Soytas et al. (2007)
1960-2004 procedure economic growth, EC→CO2
CO2 emission
ARDL bound test, Energy
18 European
Acaravci and error correction consumption,
countries, mostly Mixed results
Ozturk (2010) model, Granger economic growth,
1960-2005
causality CO2 emission
Panel ARDL bound
4 oil-exporting
Bildirici and test, error Oil production,
countries, OP↔Y
Kayıkçı (2013) correction model, economic growth
1993-2010
Granger causality
20 OECD Granger causality Energy
Bozoklu and
countries, mostly in the frequency consumption, Mixed results
Yilanci (2013)
1970-2011 domain economic growth
Energy
Cointegration, error consumption, EC→Y
Alshehry and Saudi Arabia,
correction model, economic growth,
Belloumi (2015) 1971-2010
Granger causality CO2 emission,
energy price
Granger causality Energy
Ahmed and Azam 119 countries,
in the frequency consumption, Mixed results
(2016) mostly 1960-2012
domain economic growth
Energy
G7 countries, Bootstrap panel
Mutascu (2016) consumption, Mixed results
1970-2012 Granger causality
economic growth
Energy
Cointegration, consumption,
Vietnam,
Tang et al. (2016) Toda-Yamamoto economic growth, EC→Y
1971-2011
procedure and control
variables
17 emerging
Destek and Aslan Bootstrap panel REC, NREC,
economies, Mixed results
(2017) Granger causality economic growth
1980-2012
Panel Energy
10 oil-exporting cointegration, error consumption,
Hasanov et al.
countries, correction model, economic growth, EC→Y
(2017)
1997-2014 panel Granger and control
causality variables

7
 ACCEPTED MANUSCRIPT

Energy
ARDL bound test,
consumption,
Nyasha et al. Ethiopia, error correction
economic growth, Y→EC
(2018) 1971-2013 model, Granger
and control
causality
variables

C: MENA Studies
Panel Energy
6 GCC countries,
Al-Iriani (2006) cointegration, panel consumption, Y→EC
1971-2002
Granger causality economic growth
Panel
OC↔Y,
17 MENA cointegration, error Oil consumption,
OC↔CO2,
Al-Mulali (2011) countries, correction model, economic growth,
Y↔CO2
1980-2009 panel Granger CO2 emission
causality
Energy
14 MENA consumption, Y↔EC,
Panel GMM,
Omri (2013) countries, economic growth, Y↔CO2,
causality analysis
1990-2011 CO2 emission, and EC→ CO2
control variables
Panel Energy
12 MENA
cointegration, error consumption,
Ozcan (2013) countries, Y→EC
correction model, economic growth,
1990-2008
Granger causality CO2 emission
Panel
10 MENA cointegration, error RE, NRE,
Farhani and Y―RE,
countries, correction model, economic growth,
Shahbaz (2014) Y―NRE
1980-2009 panel Granger CO2 emission
causality
Energy
Panel techniques,
Salahuddin and 6 GCC countries, consumption,
panel Granger Y→EC
Gow (2014) 1980-2012 economic growth,
causality
CO2 emission
Panel
REC, NREC,
11 MENA cointegration, panel
economic growth, Y↔REC
Kahia et al. (2017) countries, error correction
and control Y↔NREC
1980-2012 model, panel
variables
Granger causality
Note: ―, →, and ↔ denote no causality, uni-directional causality, and bi-directional causality,
respectively. Y: economic growth, EC: energy consumption, CO2: carbon dioxide emission, OP: oil
production, OC: oil consumption, RE: renewable electricity consumption, NRE: non-renewable
electricity consumption, REC: renewable energy consumption, NREC: non-renewable energy
consumption.
3. Data and Methodology

3.1. Data

In this study, we examine the causal relationship between CO2 emission, energy consumption,
and economic growth for eight MENA countries; namely, Algeria, Egypt, Iran, Iraq, Oman,
Saudi Arabia, Tunisia, and UAE for the period 1975-2014. For this aim, the study uses annual
data of GDP per capita (constant 2010 US dollars), carbon dioxide emissions per capita (metric
tons), and energy use per capita (kg of oil equivalent), and these variables are labeled as lnY,

8
 ACCEPTED MANUSCRIPT

lnCO2, and lnEC, respectively. All data are obtained from the World Bank database, and all the
series are used in natural logarithms.

3.2. Methodology

Time domain analysis evaluates the time series considered as a function of time while in the
case of frequency domain analysis, the series considered is regarded as a function of frequency
by separating it into various frequency components. However, the spectral density function is
investigated in frequency domain analyses, and therefore periodic fluctuations in the series are
revealed by frequency domain analyses. Granger causality analysis is extensively used to
examine the relationship between variables in the literature however it has some disadvantages.
First, conventional Granger causality tests calculate a single test statistic over time. Thus, such
tests cannot examine the causality relationship at different frequencies. Second, if the
relationship between variables is connected more than one frequency rather than one, the time
dimension approach is not sufficient to use the information in the original data set effectively
(Masset, 2008). However, Geweke (1982) developed the Wald test procedure, which introduced
linear constraints on coefficient parameters to test Granger causality in a certain frequency
range. This test procedure was developed by Breitung and Candelon (2006) as a single-country
test for the causality test in the frequency domain. Besides, Croux and Reusens (2013)
developed a causality test in the frequency domain for panel data. This recently developed
frequency domain test for panel data analysis computes the Granger causality for each
individual frequency component in place of calculating a single Granger causality measure for
the entire causal association. The panel Granger causality analysis in the frequency domain
makes it possible to determine whether the predictive power is concentrated at quickly
fluctuating components or at slowly fluctuating components (Croux and Reusens, 2013).

3.2.1. Single-Country Causality Test

The Vector Autoregression (VAR) models are developed by Sims (1980). Sims (1980) states
that no distinction should be made between endogenous and exogenous variables if there is
simultaneity among variables. So, all variables are regarded as endogenous when this
distinction is abandoned (Asteriou and Hall, 2011). The VAR models include the vector of
endogenous variables as an autoregressive function of their lagged terms. Also, some
exogenous variables such as trends and seasonal dummies can be included in these models

9
 ACCEPTED MANUSCRIPT

(Baltagi, 2008). Xt and Yt are stationary time series. The vector autoregression model used in
this study is as follows1:

 X t   11 ( L) 12 ( L)   X t   1t 

 ( L)       , (1)
 Yt    21 ( L)  22 ( L)   Yt    2t 
where  ( L)  I  1 ( L)   2 ( L2 )  ...   p ( Lp ) is a lag polynomial of order p with Lj X t  X t  j

and Lj X t  X t  j . 1t and  2t are error terms with white noise ( E ( t )  0 and E ( t  t' )   ).  is

symmetric and positive definite.

When this VAR equation is transformed into the moving average (MA) process; Xt is a sum of
two noncorrelated MA processes. This process involves the past shocks in Xt and the estimation
power of the Yt variable. In this case, the causality between Xt and Yt can be determined as
follows: if the predictive component of the Xt spectrum at w frequency is zero, Yt does not
Granger cause Xt at frequency w.

Breitung and Candelon (2006) proposed a single test procedure by putting the constraints on
the first VAR model:

p p
X t   11. j X t  j   12. jYt  j  1t , (2)
j 1 j 1

where 11. j and 12. j refer to the coefficients of the lag polynomials. 1t is the error term and 

is the lag length. The constraints on the null hypothesis of “no Granger causality from Yt to Xt
at the frequency w” are as follows:

j 1
12. j cos( jw)  0,

p
(3)
j 1
12. j sin( jw)  0.

These constraints can be tested with the standard F test or with the help of the incremental R 2
measured test, which is calculated as follows:

RI2  R 2  R*2 . (4)

where R 2 and R*2 are calculated from unrestricted and restricted model, respectively. If this

condition occurs, the null hypothesis is rejected.

1 See Sims (1980) for the details of the VAR models.

10
 ACCEPTED MANUSCRIPT

2
RI2  F(2T  2 p ,1 ) (1  R 2 ) . (5)
T 2p
3.2.2. Multi-Country Causality Test

The seemingly unrelated regression (SUR) model used in this study is as follows (Croux and
Reusens, 2013):

p p
X i ,t   i , j X i ,t  j    i , jYi ,t  j  ei ,t , i  1, 2,3,..., N . (6)
j 1 j 1

where X i ,t and Yi ,t refer to the variables of country i at time t. ei ,t is the error term at time t of

the country i.  is the lag length, and N is the number of countries. The constraints on the null
hypothesis are as follows:


j 1
i, j cos( jw)  0, i  1, 2,..., N .

p
(7)

j 1
i, j sin( jw)  0, i  1, 2,..., N .

These constraints are tested with the help of the incremental R 2 measured test, which is
calculated as follows:

RI2  R 2  R*2 . (8)

where R 2 is unrestricted and R*2 is the restricted McElroy R 2 value (McElroy, 1977).

2N
RI2  F(2 N , N (T  2 p ),1 ) (1  R 2 ) (9)
N (T  2 p )
If this condition occurs, the null hypothesis of “no Granger causality exists from Yt to Xt at the
frequency w in all countries” is rejected.

4. Empirical Results

In this section, we present the empirical results of the study. First, this study investigates cross-
section dependency (CD) of the series and then check the stationarity properties of the variables
through panel unit root tests. In this study, we use three cross-section dependence tests; namely,
CDBP, CDLM, and CD. Table 2 shows the results of both the cross-section dependence tests and
panel unit root test. Empirical findings demonstrate that the cross-sectional dependency exists
in the series. In other words, the impact of a shock from any of the countries in the panel also
affects other units. Thus, this study employs a second generation panel unit root test, CADF

11
 ACCEPTED MANUSCRIPT

(Cross-Sectionally Augmented Dickey-Fuller), developed by Pesaran (2007) which considers

the cross-section dependence of the series. Table 2 also displays the stationarity properties of
all the variables. The empirical results show that lnY is stationary at the first difference, I(1),
while lnEC and lnCO2 are stationary at their level value, I(0).

Table 2. Cross-Section Dependence and Panel Unit Root Tests for The Series
Variables CDBP CDLM CD CIPS Statistics
lnY 447.952*** 56.118*** 3.149*** -0.943
lnEC 678.169*** 86.882*** 25.226*** -2.341**
lnCO2 358.746*** 44.197*** 9.165*** -2.339**
∆lnY - - - -3.846***
Note: *** and ** indicate the rejection of the null hypothesis at the 1% and 5% significance level,
respectively. CIPS statistics is the simple average of the individual CADF statistics (𝐶𝐴𝐷𝐹𝑖).

In this study, the directions of the causal relationship between variables are investigated at
frequencies of 2.5, 1.5, and 0.5. These frequencies denote the short-, the intermediate-, and the
long-run causal relationships, respectively. In other words, 0.5 shows permanent causality while
2.5 displays temporary causality.

Table 3 shows the frequency domain causality test results between economic growth and CO2
emissions for each country. According to the results, there is a uni-directional temporary
causality from economic growth to CO2 emissions in Algeria. In addition, there is a one-way
causality running from economic growth to emission level in the intermediate-run in Oman. On
the other hand, there is a permanent causality from CO2 emissions to economic growth in the
case of Oman. In most of the cases, there is no link between these variables. Thus, one can state
that the emission-curbing policies are applicable in these countries. However, it is hard to
conduct this type of policies in Oman due to the existence of the one-way causality from
emission level to economic growth in the long-run.

Table 3. Granger Causality Analysis in the Frequency Domain Results (lnY, lnCO2)
H0: lnY 
 lnCO2 H0: lnCO2 
 lnY
Countries w=0.5 w=1.5 w=2.5 c.v.=10% w=0.5 w=1.5 w=2.5 c.v.=10%
Algeria 0.031 0.087 0.129* 0.110 0.036 0.038 0.020 0.114
Egypt 0.001 0.004 0.006 0.008 0.036 0.028 0.026 0.133
Iran 0.006 0.002 0.001 0.008 0.022 0.015 0.017 0.116
Iraq 0.052 0.022 0.024 0.081 0.088 0.070 0.108 0.121
Oman 0.003 0.017* 0.007 0.009 0.135* 0.001 0.006 0.117
Saudi Arabia 0.021 0.004 0.006 0.072 0.049 0.035 0.030 0.129
Tunisia 0.001 0.003 0.005 0.012 0.012 0.023 0.024 0.161

12
 ACCEPTED MANUSCRIPT

UAE 0.031 0.004 0.024 0.079 0.013 0.024 0.016 0.142

Note: indicates the rejection of the null hypothesis at 10% significance level.
*

Table 4 presents the causal nexus between energy consumption and economic growth for eight
oil-rich MENA countries in terms of single-country results. In the short-run, there is a uni-
directional causality from economic growth to energy consumption which supports the
conservation hypothesis in Algeria, Iran, Iraq, Tunisia, and UAE while there is a bi-directional
relationship that confirms the feedback hypothesis in Saudi Arabia. In addition, a one-way
causality exists from economic growth to energy use in Iran, Saudi Arabia, Tunisia, and UAE
in the intermediate-run. Lastly, a uni-directional permanent Granger causality runs from lnY to
lnEC in Iran, Iraq, Saudi Arabia, and UAE. On the other side, there is only a one-way causal
nexus from lnEC to lnY that confirms the growth hypothesis in Oman in the long-run. Thus,
one can reveal that the policymakers can apply energy conservation policies except for Oman
in the long-run and Saudi Arabia in the short-run, easily.

Table 4. Granger Causality Analysis in the Frequency Domain Results (lnY, lnEC)
H0: lnY 
 lnEC H0: lnEC 
 lnY
Countries w=0.5 w=1.5 w=2.5 c.v.=10% w=0.5 w=1.5 w=2.5 c.v.=10%
Algeria 0.007 0.009 0.016* 0.014 0.036 0.039 0.042 0.111
Egypt 0.000 0.001 0.002 0.004 0.030 0.058 0.054 0.126
Iran 0.005 * 0.006 * 0.005 * 0.004 0.078 0.067 0.075 0.099
Iraq 0.053* 0.027 0.031* 0.028 0.021 0.026 0.031 0.135
Oman 0.000 0.002 0.001 0.011 0.191 * 0.078 0.073 0.108
Saudi Arabia 0.032 * 0.025 * 0.015 * 0.013 0.088 0.116 0.119* 0.117
Tunisia 0.001 0.005* 0.006* 0.004 0.010 0.100 0.098 0.148
UAE 0.039 * 0.038 * 0.037 * 0.020 0.011 0.014 0.017 0.143
Note: indicates the rejection of the null hypothesis at 10% significance level.
*

Table 5 gives the single-country Granger causality test analysis results in the frequency domain
between energy consumption and CO2 emissions in the short-, the intermediate-, and the long-
run. According to the empirical findings, there is a uni-directional causality from energy
consumption to emission level in Egypt and Tunisia while a one-way causality exists from
lnCO2 to lnEC in Oman in the short-run. Besides, the directions of the causation are same as
the short-run analysis in the intermediate-run for all the countries in the sample. We can reveal
that energy consumption is the Granger cause of carbon dioxide level in Algeria, Egypt, Iran,
Oman, and Tunisia when the long-run findings are considered. It is seen that there is a pollutive
effect of energy consumption in the long-term.

Table 5. Granger Causality Analysis in the Frequency Domain Results (lnEC, lnCO2)

13
 ACCEPTED MANUSCRIPT

H0: lnEC 
 lnCO2 H0: lnCO2 
 lnEC
Countries w=0.5 w=1.5 w=2.5 c.v.=10% w=0.5 w=1.5 w=2.5 c.v.=10%
Algeria 0.250 * 0.053 0.052 0.090 0.020 0.001 0.004 0.023
Egypt 0.011* 0.013* 0.014* 0.006 0.005 0.006 0.006 0.016
Iran 0.024 * 0.004 0.001 0.005 0.001 0.002 0.003 0.005
Iraq 0.030 0.012 0.018 0.085 0.010 0.020 0.008 0.035
Oman 0.018 * 0.002 0.003 0.009 0.002 0.013* 0.015* 0.009
Saudi Arabia 0.001 0.002 0.004 0.070 0.007 0.011 0.012 0.017
Tunisia 0.021 * 0.011 * 0.013 * 0.009 0.002 0.003 0.002 0.005
UAE 0.001 0.003 0.004 0.085 0.003 0.001 0.002 0.004
Note: indicates the rejection of the null hypothesis at 10% significance level.
*

Single-country Granger causality in the frequency domain results considering lnY-lnCO2,

lnY-lnEC, and lnEC-lnCO2 relationships provide mixed results for each country. Policymakers
of the countries can determine national energy and environment policies across different
frequencies, separately. Countries in which the conservation and the neutrality hypotheses are
confirmed can manage the domestic energy demand without damaging economic growth easier
than other countries. Besides, restricting (decreasing) energy use rises environmental quality in
these countries. All of these are oil-rich MENA countries, but their income levels are not same.
Differences in income levels might be the main reason behind the results differing between
MENA countries. For instance, GDP per capita (constant 2010 US dollars) in 2014 is $39,034
in UAE; $21,183 in Saudi Arabia; and $17,132 in Oman while the income levels of the
remaining countries are below $7,000 according to the World Bank data. Besides, some of these
countries complete their transformation to service (post-industrial) economy while others are
still at the industrial stage. According to the Environmental Kuznets Curve, after a threshold
level of income, environmental degradation decreases when income rises. Thus, people are
more aware of environmental issues in rich countries than in developing countries because the
environment can be regarded as a luxury in emerging economies. Moreover, government
policies on energy and environment affect the behaviors of energy use of the economic agents
such as households, firms, organizations, etc. These reasons lead to differences in empirical
findings for each country across different frequencies.

Table 6 displays the results of panel Granger causality analysis in the frequency domain
considering eight oil-rich MENA countries as a whole. Also, an overview of the panel Granger
causality in the frequency domain for economic growth, CO2 emission, and energy consumption
is displayed in Figure 1. Empirical findings demonstrate that there is no causal nexus between
economic growth and emission level in the short-, the intermediate-, and the long-run. Also,

14
 ACCEPTED MANUSCRIPT

there is a one-way causality running from economic growth to energy use both in the short- and
the intermediate-run, and a uni-directional causal relationship exists from lnEC to lnY in the
long-run. So, one can infer that the conservation hypothesis is valid in the short- and the
intermediate-run while the growth hypothesis is valid in the long-run. Lastly, the causal nexus
between energy consumption and pollution level is examined. According to empirical findings,
energy consumption is the Granger cause of emission level at all frequencies. So, it is clear that
the past values of the energy usage can be used to estimate future values of the emission level.
Moreover, one can also use the past values of emission level to predict future values of energy
consumption both in the intermediate- and the long-run. Shortly, there is a bi-directional
relationship between lnEC and lnCO2 in the intermediate- and the long-run while there is only
a uni-directional causality running from lnEC to lnCO2 in the short-run.

Table 6. Panel Granger Causality Analysis in the Frequency Domain Results

The Null Hypothesis w=0.5 w=1.5 w=2.5 c.v.=10%

lnY  lnCO2 0.004 0.008 0.003 0.013
lnCO2  lnY 0.074 0.040 0.030 0.075
lnY  lnEC 0.004 0.012 * 0.010* 0.005
lnEC  lnY 0.094* 0.041 0.044 0.072
lnEC  lnCO2 0.071 * 0.044 * 0.045* 0.006
lnCO2  lnEC 0.007* 0.008* 0.005 0.006
Note: indicates the rejection of the null hypothesis at 10% significance level.
*

[Figure 1 near here]

The study also conducts a panel causality test developed by Dumitrescu and Hurlin (2012) for
heterogenous panels in the time domain to compare the results with the panel causality test in
the frequency domain. Table 7 displays the Dumitrescu-Hurlin causality test results that
examine the null of no causal relationship for any of the cross-section units against the
alternative of at least one causal relationship exists in the cross-section units (Dumitrescu and
Hurlin, 2012). This study reports Zbar-statistics of the test for both one lag and two lags since
there is no criteria for choosing an optimal lag length for Dumitrescu-Hurlin causality test.
According to the results, economic growth is the Granger cause of energy consumption in the
MENA countries. Besides, there is a two-way causal relationship between energy consumption
and economic CO2 emissions. After all, one can state that the conservation hypothesis is valid

15
 ACCEPTED MANUSCRIPT

in the MENA countries. Also, energy consumption and emissions are jointly determined and
affected at the same time.

Table 7. Dumitrescu-Hurlin Panel Causality Test Results

Zbar-Statistics Zbar-Statistics
The Null Hypothesis Conclusion
(for 1 lag) (for 2 lags)
lnY 
 lnCO2 0.803 0.118 Do not reject the null.
lnCO2  lnY 0.736 0.435 Do not reject the null.
lnY 
 lnEC 2.526** 3.382*** Reject the null.
lnEC  lnY -0.355 0.747 Do not reject the null.
lnEC  lnCO2 5.796*** 4.940*** Reject the null.
lnCO2  lnEC 1.686* 2.708*** Reject the null.
Note: ***, **, and * indicate the rejection of the null hypothesis at the 1%, 5%, and 10% significance
level, respectively.

The empirical results of the Panel Granger causality analyses both in the time domain and in
the frequency domain are reported in Table 6 and Table 7. The Dumitrescu-Hurlin causality
test finds only one uni-directional causality running from economic growth to energy
consumption and one bi-directional causality between energy consumption and CO2 emissions
while the frequency domain analysis detects more causal nexus between variables across
different frequencies. Thus, these results may provide better information about the relationship
between energy consumption, economic growth, and CO2 emissions to the policymakers for
the short-, the intermediate-, and the long-run.

To our best knowledge, there is not any study on energy-growth-emission nexus in the literature
that employs panel Granger causality test in the frequency domain. It is not rational to compare
our empirical results with the findings of the previous studies because the existing literature
mostly considers the causal nexus between aforementioned variables in the time domain. Since
the methodology of this study can detect causal relationships across different frequencies that
the conventional Granger causality analysis cannot find, our findings provide more reliable and
valuable information to the policymakers.

5. Conclusion

This study examines the causal relationship between energy consumption, economic growth,
and CO2 emission through both single- and multi-country Granger causality analysis in the
frequency domain in eight oil-rich MENA countries during the period 1975-2014. Single-
country results suggest that the causal nexus between variables are mixed across different

16
 ACCEPTED MANUSCRIPT

frequencies. Policymakers of these countries may take into consideration these empirical
findings to design effective and proper energy and environmental conservation policies in the
short-, the intermediate-, and the long-run.

According to the results of the panel Granger causality analysis in the frequency domain, there
is no causal nexus between economic growth and emission level in this set of countries. Thus,
one can state that an environmental conservation policy does not affect economic growth
negatively. Moreover, the energy conservation hypothesis holds in the short- and the
intermediate-run while the growth hypothesis is valid in the long-run in the MENA countries.
So, even if the energy conservation policies are effective in the short- and the intermediate-run,
their impact on the economic growth will be negative in the long-term. Lastly, this study finds
that there is a feedback effect between energy consumption and emission level in the
intermediate- and the long-run. However, a uni-directional Granger causality is confirmed from
energy consumption to emission level in the short-run. The empirical findings of this study may
provide beneficial information to policymakers of the region to shape effective conservation
policies about energy and environment such as reducing energy demand and controlling
emission level across different frequencies.

MENA countries have faced crucial environmental problems for a long time, and it is expected
that these issues will continue if necessary measures are not taken. Further studies on this
subject can be expanded by examining more countries in the MENA region. Besides, samples
can be divided into country groups according to the macro or micro characteristics of the
economies. In addition, renewable and non-renewable energy consumption data can be used
instead of total energy use data. Also, using aggregated and disaggregated energy consumption
data may lead more beneficial results to the policymakers of the region. Lastly, to conduct
recently developed econometric techniques may increase the reliability of the empirical
findings.

References

Acaravci, A., & Ozturk, I. (2010). On the relationship between energy consumption, CO2
emissions and economic growth in Europe. Energy, 35(12), 5412-5420.

17
 ACCEPTED MANUSCRIPT

Ahmed, M., & Azam, M. (2016). Causal nexus between energy consumption and economic
growth for high, middle and low income countries using frequency domain
analysis. Renewable and Sustainable Energy Reviews, 60, 653-678.

Akarca, A. T., & Long, T. V. (1980). On the relationship between energy and GNP: A
reexamination. The Journal of Energy and Development, 5, 326-331.

Al-Iriani, M. A. (2006). Energy–GDP relationship revisited: An example from GCC countries

using panel causality. Energy policy, 34(17), 3342-3350.

Al-Mulali, U. (2011). Oil consumption, CO2 emission and economic growth in MENA
countries. Energy, 36(10), 6165-6171.

Alshehry, A. S., & Belloumi, M. (2015). Energy consumption, carbon dioxide emissions and
economic growth: The case of Saudi Arabia. Renewable and Sustainable Energy
Reviews, 41, 237-247.

Asteriou, D., & Hall, S. G. (2011). Applied Econometrics. Palgrave Macmillan.

Baltagi, B. H. (2008). Econometrics. Springer.

Bildirici, M. E., & Kayıkçı, F. (2013). Effects of oil production on economic growth in Eurasian
countries: Panel ARDL approach. Energy, 49, 156-161.

Bozoklu, S., & Yilanci, V. (2013). Energy consumption and economic growth for selected
OECD countries: Further evidence from the Granger causality test in the frequency
domain. Energy Policy, 63, 877-881.

Breitung, J., & Candelon, B. (2006). Testing for short- and long-run causality: A frequency-
domain approach. Journal of Econometrics, 132(2), 363-378.

British Petroleum. (2017). BP statistical review of world energy 2017.

Copeland, B. R. (2010). The Pollution Haven Hypothesis. In Gallagher, K. (Ed.), Handbook on

Trade and The Environment. Edward Elgar Publishing.

Croux, C., & Reusens, P. (2013). Do stock prices contain predictive power for the future
economic activity? A Granger causality analysis in the frequency domain. Journal of
Macroeconomics, 35, 93-103.

18
 ACCEPTED MANUSCRIPT

Destek, M. A., & Aslan, A. (2017). Renewable and non-renewable energy consumption and
economic growth in emerging economies: Evidence from bootstrap panel causality.
Renewable Energy, 111, 757-763.

Dumitrescu, E. I., & Hurlin, C. (2012). Testing for Granger non-causality in heterogeneous
panels. Economic Modelling, 29(4), 1450-1460.

Erol, U., & Yu, E. S. H. (1987). On the causal relationship between energy and income for
industrialized countries. The Journal of Energy and Development, 13(1), 113-122.

Farhani, S., & Shahbaz, M. (2014). What role of renewable and non-renewable electricity
consumption and output is needed to initially mitigate CO2 emissions in MENA
region?. Renewable and Sustainable Energy Reviews, 40, 80-90.

Geweke, J. (1982). Measurement of linear dependence and feedback between multiple time
series. Journal of the American Statistical Association, 77(378), 304-313.

Grossman, G. M., & Krueger, A. B. (1991). Environmental impacts of a North American Free
Trade Agreement (No. w3914). National Bureau of Economic Research.

Hasanov, F., Bulut, C., & Suleymanov, E. (2017). Review of energy-growth nexus: A panel
analysis for ten Eurasian oil exporting countries. Renewable and Sustainable Energy
Reviews, 73, 369-386.

Hwang, D. B., & Gum, B. (1991). The causal relationship between energy and GNP: The case
of Taiwan. The Journal of Energy and Development, 16(2), 219-226.

Kahia, M., Aïssa, M. S. B., & Lanouar, C. (2017). Renewable and non-renewable energy use-
economic growth nexus: The case of MENA Net Oil Importing Countries. Renewable
and Sustainable Energy Reviews, 71, 127-140.

Karanfil, F. (2009). How many times again will we examine the energy-income nexus using a
limited range of traditional econometric tools?. Energy Policy, 37(4), 1191-1194.

Kraft, J., & Kraft, A. (1978). On the relationship between energy and GNP. The Journal of
Energy and Development, 3(2), 401-403.

Masset, P. (2008). Analysis of financial time-series using Fourier and wavelet methods.
Fribourg: University of Fribourg.

19
 ACCEPTED MANUSCRIPT

McElroy, M. B. (1977). Goodness of fit for seemingly unrelated regressions: Glahn's R2y.x and
Hooper's r2. Journal of Econometrics, 6(3), 381-387.

Mehrara, M. (2007). Energy consumption and economic growth: the case of oil exporting
countries. Energy policy, 35(5), 2939-2945.

Mutascu, M. (2016). A bootstrap panel Granger causality analysis of energy consumption and
economic growth in the G7 countries. Renewable and Sustainable Energy Reviews, 63,
166-171.

Nachane, D. M., Nadkarni, R. M., & Karnik, A. V. (1988). Co-integration and causality testing
of the energy–GDP relationship: A cross-country study. Applied Economics, 20(11),
1511-1531.

Nyasha, S., Gwenhure, Y., & Odhiambo, N. M. (2018). Energy consumption and economic
growth in Ethiopia: A dynamic causal linkage. Energy & Environment. Advance online
publication. doi: 10.1177/0958305X18779574

Omri, A. (2013). CO2 emissions, energy consumption and economic growth nexus in MENA
countries: Evidence from simultaneous equations models. Energy Economics, 40, 657-
664.

Ozcan, B. (2013). The nexus between carbon emissions, energy consumption and economic
growth in Middle East countries: a panel data analysis. Energy Policy, 62, 1138-1147.

Ozturk, I. (2010). A literature survey on energy–growth nexus. Energy Policy, 38(1), 340-349.

Pesaran, M. H. (2007). A simple panel unit root test in the presence of cross‐section
dependence. Journal of Applied Econometrics, 22(2), 265-312.

Salahuddin, M., & Gow, J. (2014). Economic growth, energy consumption and CO2 emissions
in Gulf Cooperation Council countries. Energy, 73, 44-58.

Sims, C. A. (1980). Macroeconomics and reality. Econometrica: Journal of the Econometric

Society, 48(1), 1-48.

Soytas, U., Sari, R., & Ewing, B. T. (2007). Energy consumption, income, and carbon emissions
in the United States. Ecological Economics, 62(3-4), 482-489.

Tang, C. F., Tan, B. W., & Ozturk, I. (2016). Energy consumption and economic growth in
Vietnam. Renewable and Sustainable Energy Reviews, 54, 1506-1514.

20
 ACCEPTED MANUSCRIPT

World Bank. (2008). Global Monitoring Report 2008 – MDGs and the environment: Agenda
for inclusive and sustainable development, Washington, DC: World Bank.

Yu, E. S. H., & Choi, J. Y. (1985). The causal relationship between energy and GNP: An
international comparison. The Journal of Energy and Development, 10(2), 249-272.

Yu, E. S. H., & Hwang, B. K. (1984). The relationship between energy and GNP: Further
results. Energy Economics, 6(3), 186-190.

21
 ACCEPTED MANUSCRIPT

lnY

lnCO2 lnEC

Note: → and ↔ denote uni-directional and bi-directional causality, respectively. Straight lines, dashed
lines, and bold dashed lines show the short-, the intermediate-, and the long-run relationships in order.
Figure 1. Energy-Emission-Growth Nexus for the MENA countries
 ACCEPTED MANUSCRIPT

Highlights

No causal nexus between economic growth and CO2 in the MENA region.

The energy conservation hypothesis holds in the short- and the intermediate-run.

Energy consumption causes economic growth in the long-run.

Energy consumption is the Granger cause of CO2 in the short-run.

Two-way causality exists between energy use and CO2 except for the short-run.