The relationship among oil and coal consumption, carbon dioxide emissions, and

economic growth in BRICTS countries

, ,
Elif Melike Bildirici and Tahsin Bakirtas

Citation: J. Renewable Sustainable Energy 8, 045903 (2016); doi: 10.1063/1.4955090

View online: http://dx.doi.org/10.1063/1.4955090
View Table of Contents: http://aip.scitation.org/toc/rse/8/4
Published by the American Institute of Physics
 JOURNAL OF RENEWABLE AND SUSTAINABLE ENERGY 8, 045903 (2016)

The relationship among oil and coal consumption, carbon

dioxide emissions, and economic growth in BRICTS
countries
Elif Melike Bildirici1,a) and Tahsin Bakirtas2,b)
1
Social Science Institute, Yıldız Technical University, Istanbul, Turkey
2
Department of Economics, Sakarya University, Sakarya, Turkey
(Received 3 January 2016; accepted 15 June 2016; published online 11 July 2016)

This study aims to analyze the relationship between carbon dioxide (CO2)
emissions, economic growth, and coal and oil consumption in Brazil, Russia, India,
China, Turkey, and South Africa by using the bounds test approach autoregressive
distributed lag (ARDL) over the period from 1969 to 2011. According to ARDL
analysis results, it is determined short-run and long-run relationships among
selected variables. Three long-run estimators: ARDL cointegration, dynamic ordi-
nary least squares, and fully modified ordinary least squares are utilized to test the
robustness of the estimation results. The Granger causality and the forecast error
variance decomposition approaches indicate the evidence of a causal relation
between variables. According to empirical results, there are the evidence of a uni-
directional Granger causality from real gross domestic product (GDP) to carbon
dioxide (CO2) emissions in analyzed countries, uni-directional causality from coal
consumption to carbon dioxide (CO2) emissions, uni-directional causality from oil
consumption to carbon dioxide (CO2) emissions in China, India, Turkey, and South
Africa and bi-directional causality in Brazil and Russia. Meanwhile, there is bidir-
ectional causality from GDP to coal consumption, from coal consumption to oil
consumption for Brazil, Russia, China, Turkey, and South Africa. India’s causality
results reveal a bidirectional causality from GDP to coal consumption, oil con-
sumption, and carbon dioxide (CO2) emissions, and the results of forecast error var-
iance supported the results of the causality test. Published by AIP Publishing.
[http://dx.doi.org/10.1063/1.4955090]

I. INTRODUCTION
Energy has played a central role in the development of countries and functioning of the
world’s economy. Because it is an essential input to agricultural production, transportation, indus-
try, and the home, reliance on energy will continue to grow as world population increases and as
economic growth and development continue. As an effect of increased mobility and telecommuni-
cations, urbanization and an integrated global economy will further accelerate our energy con-
sumption and dependence. History has shown us that increased energy consumption brings its
own burdens with respect to the environment, health, safety, lifestyle, and community (Connors,
1998). Thus, global warming and climate change caused by fossil energy consumption have been
two of the most significant problems for the environment since the mid-20th century. From an
economic viewpoint, the energy sector has been responsible for the emergence of significant envi-
ronmental pollution caused by energy consumption and transmission. An increasing amount of
carbon dioxide (CO2) emissions has been an important cause of environmental pollution.
Academic studies and international institutions have concern about the impacts of “global
warming” on rainfall patterns, sea level, the frequency of severe storms, ecosystems (Connors,

a)
bildiri@yildiz.edu.tr
b)
tahsinbakirtas@gmail.com

1941-7012/2016/8(4)/045903/15/$30.00 8, 045903-1 Published by AIP Publishing.

045903-2 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

1998), and environmental pollution. Studies on environmental pollution have claimed its close
relationship to an environmental Kuznets curve (EKC) shaped as both an inverted U and N. On
the other hand, the relationship between economic growth and environmental pollution has
been widely considered. From one aspect, the environmental Kuznets’ curve (EKC) hypothesis
assumes that an inverted U-shaped relationship exists between environmental pollution and
income per capita. According to this view, environmental degradation increases alongside
income per capita during earlier stages of economic growth, after arriving at a threshold value
it starts to decline while income per capita continues to grow.
The national economies analyzed in this paper, especially China, India, and Turkey, have
continued to grow despite the recent recession. As a result, primary energy consumption needs
to expand at an average annual rate of 7.4% in China and 5.1% in India, both of which are
much faster than required by the rest of the world. China and India together showed a signifi-
cant increase in their share of total world energy consumption: from 10% in 1990 to 27.1% in
2013. According to British Petroleum, in 2013 emerging economies accounted for 80% of the
global increase in energy consumption and the BRICTS countries—Brazil, Russia, India, China,
Turkey, and South Africa—consumed 36.8% of the world’s primary energy (24.7% of the
world’s total oil, 21.2% natural gas, and 64.6% coal) and produced 41.3% of the world’s CO2
emissions. China alone consumed 21.9% of the world’s primary energy (50.3% of the world’s
total coal, 12.1% oil, 24.1% hydroelectricity, 4.8% natural gas, and 4.4% nuclear energy) while
producing 27.1% of the world’s total CO2 emissions (BP, 2014).
Coal and oil consumption contributes more CO2 emissions than other energy sources, and
consumption of these energy sources that plays a primary role in economic growth is one of
the largest sources of the world’s total CO2 emissions. Although reducing coal and oil energy
consumption seems to be an important factor in mitigating CO2 emissions, the importance of
its impact on economic development complicates the reduction of such consumption
(Govindaraju and Tang, 2013).
Understanding the relationship between coal and oil consumption, CO2 emissions, and eco-
nomic development in these countries is very important. This study aims to analyze links
between CO2 emissions, economic growth, and coal and oil consumption in the countries that
exhibited high economic development. Although this study can be considered to complement
previous empirical studies, it differs from the existing literature in some aspects. First, it ana-
lyzes BRICTS countries (BRICTS concept was used by Bildirici and Bakırtas (2014)) in order
to discover the relationship between CO2 emissions, economic growth, and coal and oil con-
sumption. Second, it is distinguished from previous work in its employment of autoregressive
distributed lag (ARDL) cointegration, the dynamic ordinary least squares (DOLS) developed by
Saikkonen (1991) and Stock and Watson (1993) and the fully modified ordinary least squares
(FMOLS) suggested by Phillips and Hansen (1990) for these countries. The reason for using
these methods is to test the robustness of the estimation results and also to provide more effi-
cient results in a small sample study. Since the long and short-run relationships between the
variables are important for BRICTS’ energy policy and strategy, we used several tests in order
to ensure the results of the paper not being restricted by only one approach. We paid special
attention to compare the results of the tests in this paper. Finally, in order to test the strength of
our causality test, we used the generalized forecast error variance decomposition (VDC) analy-
sis developed by Pesaran and Shin (1998), which is invariant to the ordering of the series.
Section II of this study reviews the literature on economic growth, coal and oil consump-
tion, and CO2 consumption. Section III introduces econometric theory and methodology while
Section IV presents empirical results. The study concludes with a discussion of the implications
for policy.

II. LITERATURE REVIEW

There are a lot of papers examining the relationship between economic growth and envi-
ronmental pollution. The main hypothesis in this literature is environmental Kuznets’ curve
(EKC) hypothesis which implies that the relationship between economic growth and
045903-3 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

environmental deterioration is an inverted U-shaped. Grossman and Krueger (1991) and Shafik
and Bandyopadhyay (1992) are the first papers reporting the inverted U-shaped EKC. Selden
and Song (1994) determined the validity of the EKC hypothesis between CO2 emissions and
gross domestic product (GDP) for 130 countries by panel data analysis for the period of
1951–1986. In the same year, Shafik (1994) analyzed the relationship between gross domestic
product (GDP) and environmental pollution of 149 countries over the period 1960–1990.
Tucker (1995) analyzed the EKC hypothesis for 137 countries for the period from 1971 to
1991 by using the panel data method. De Bruyn et al. (1998) tested the relationship between
environmental pollution and income in four countries—namely, the UK, the US, Western
Germany, and the Netherlands—during the period from 1961 to 1993. Meanwhile, Cole et al.
(1997) tested the validity of the EKC hypothesis for countries of the Organisation for
Economic Cooperation and Development (OECD) for the period of 1970–1992. In pursuit of
these pioneering studies, several other papers analyzed the validity of the EKC hypothesis.
The research papers investigating the relationship between pollutant emissions, energy con-
sumption, and economic growth with econometric models have been popular in recent years.
However, many studies have analyzed the relationships between energy consumption, economic
growth, and pollution by combining literature on EKC such as Richmond and Kaufman (2006),
Soytas et al. (2007), Ang (2007), Soytas and Sari (2009), Akbostanci et al. (2009), Jalil and
Mahmud (2009), Zhang and Cheng (2009), Sanjari and Delangizan (2010), Shahbaz et al.
(2010), Apergis and Payne (2010), Fodha and Zaghdoud (2010), Menyah and Wolde-Rufael
(2010), Arouri et al. (2011), Wang et al. (2011), Pao and Tsai (2011), and Alam et al. (2011).
Table I summarizes the literature addressing the relationship between coal and oil con-
sumption, CO2 emissions, and economic growth.
In this paper, we aim to analyze links between CO2 emissions, economic growth, and coal
and oil consumption in the countries. This study aims to complement previous empirical stud-
ies. Understanding the relationship between coal and oil consumption, CO2 emissions, and eco-
nomic development in these countries that in the countries that exhibited high economic devel-
opment is very important.

III. DATA SPECIFICATIONS AND ECONOMETRIC METHODOLOGY

A. Data specifications
The annual data used in this study cover the period from 1969 to 2011 for Brazil, Russia,
India, China, Turkey, and South Africa (BRICTS). Selected variables are oil energy consumption
(PC), coal energy consumption (CC), CO2 emission (COC), and real GDP (Y). Oil consumption is
measured in millions of tonnes; coal energy consumption is expressed in terms of million of tonnes
oil equivalent, and CO2 emissions are measured in millions of tonnes and real GDP is expressed
in terms 2005, $billions. The real GDP data was taken from the World Bank Database, other data
were taken from the BP Statistical Review Database. All variables were converted to the logarith-
mic form to minimize skewness. Graphical representation of data is given before analysis to see
general outlook of countries. Fig. 1 presents real GDP series of BRICTS countries. Figures 2 and
3 show coal and oil consumption, respectively. According to the figures, it is seen easily that
China has the highest consumer of coal and oil. Fig. 4 reveals that China has the highest carbon
dioxide emission level and distant from the other countries which are close to each other.

B. Econometric methodology
1. ARDL method
The ARDL method has some advantages according to other conventional cointegration
tests. Emran et al. (2007) determined that the ARDL bounds tests have some advantage,
depending upon Monte Carlo evidence. The ARDL method corrects for possible endogeneity of
explanatory variables (Wolde-Rufael, 2010a; 2010b). A further added advantage of the ARDL
method is that it can be applied to studies that have a small sample size (Narayan, 2005).
However, the ARDL procedure can be applied if the regressors are I(1) and/or I(0) while
TABLE I. Summary of selected literature and major findings.

045903-4
Authors Country and time period Methodology Variables Causality

Soytas et al. (2007) United States (1960–2004) Granger causality, Energy, GDP, CO2 E ! CO2
Toda-Yamamoto (TY) approach,

E. M. Bildirici and T. Bakirtas

Generalized Variance
Decomposition (VDC) and
Impulse Response (IR)
Soytas and Sari (2009) Turkey (1960–2000) Granger causality, Energy, GDP, CO2 labor, CO2 ! E
Toda-Yamamoto (TY) fixed capital investment
approach, Generalized variance
and Impulse Response (IR)
Akbostanci et al. (2009) Turkey (1968–2003) Time series analysis, cointegration GDP, CO2 A monotonically increasing
relationship between CO2 and
income is found in the long-run
Ang (2008) Malaysia (1970–1999) Johansen cointegration, CO2, energy, GDP Strong causality: GDP !
vector error correction E (short and long run).
model, causality tests Weak causality: CO2 ! GDP
Ang (2007) France (1960–2000) Cointegration, vector CO2, energy, GDP SR (short-run): E ! GDP
correction model, Granger causality

J. Renewable Sustainable Energy 8, 045903 (2016)

Pao and Tsai (2011) Brazil, Russian Panel cointegration CO2, energy, GDP,FDI SR: E ! CO2; E $ GDP
Federation, India, LR (long-run): GDP ! CO2; GDP ! E
and China(1992–2007)
Wang et al. (2011) China (1995–2007) Pedroni cointegration; CO2, energy, GDP SR; E $ GDP, CO2 $ ELR,
Granger causality E $ CO2; GDP ! CO2
Halicioglu (2009) Turkey (1960–2005) EKC hypothesis, Granger causality GDP, energy foreign trade, CO2 E $ CO2
Li and Leung (2012) China (1985–2008) Pedroni cointegration; Granger causality Coal, GDP C $ GDP (costal & eastern region)
GDP ! C (western region)
Bloch et al. (2012) China (1965–2008) Johansen–Juselius; Granger causality DD: coal, GDP, price, CO2 SS: DD: C ! GDP; C $ CO2; SS: C ! GDP
GDP, labour, capital, coal
Chang (2010) China (1981–2006) Multivariate cointegration Energy, GDP CO2, coal, GDP ! CO2; GDP ! C,
test, vector error correction crude oil and electricity crude oil and electricity
Yuan et al. (2008) China (1963–2005) Johansen’s test; Granger causality Coal, GDP, labour, capital GDP ! C
Jinke et al. (2008) China, S.Africa, India (1980–2005) Johansen–Juselius; Coal, GDP, CO2 GDP!C(China);
Granger causality C 6¼ GDP(S.Africa, India)
 045903-5
TABLE I. (Continued.)

Authors Country and time period Methodology Variables Causality

Wolde-Rufael China, India(1965–2005) Toda-Yamamoto-Wald test causality Coal, GDP, labour, capital GDP ! C(China) C ! GDP(India)

E. M. Bildirici and T. Bakirtas

(2010a; 2010b)
Jalil and Mahmud (2009) China (1975–2005) ARDL; Granger causality GDP, CO2 GDP ! CO2
Zhang and Cheng (2009) China (1960–2007) Toda-Yamamoto-Wald test causality GDP, energy, CO2 GDP ! E, GDP 6¼ CO2
Alam et al. (2011) India (1971–2006) Toda-Yamamoto-Wald test causality Energy, GDP, CO2, labour, capital E $ CO2; E 6¼ GDP
Menyah and South Africa (1965–2006) Cointegration, Energy, GDP, CO2, pollutant emissions !
Wolde-Rufael (2010) Granger causality labour, capital, labour and capital GDP; E ! GDP; E ! CO2
Kulshreshtha India (1970–1995) Johansen–Juselius; Granger causality Coal, GDP …
and Parikh (2000)
Paul and Bhattacharya India (1950–1996) Engle–Granger; Johansen–Juselius; Energy, GDP E $ GDP
(2004) Granger causality
Ghosh (2010) India (1971–2006) ARDL; Johansen–Juselius CO2, GDP CO2 $ GDP; E ! CO2
Govindaraju India, China (1965–2009) Granger causality test CO2, GDP, coal consumption LR: GDP ! CO2 (China);
and Tang (2013) SR GDP $ CO2 (India);
CO2 $ C(India)

J. Renewable Sustainable Energy 8, 045903 (2016)

045903-6 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

FIG. 1. Real GDP (2005 $billions).

conventional cointegration technique requires that all variables in the system be of the equal
order of integration. Another important advantage is that as the conventional cointegration
method estimates the long-run relationships within a context of a system of equations, the
ARDL method employs only a single reduced form equation. The ARDL method avoids the
larger number of the specification to be made in the standard cointegration test. The last
advantage is that both the short- and the long-run relationships by the ARDL method are per-
mitted simultaneously by the estimation.
The ARDL model for the standard log-linear functional specification of a long-term rela-
tionship between variables with an OLS estimation technique is as follows:

X
m X
n X
k
D ln Yt ¼ a0 þ bi D ln Yti þ /i D ln PCti þ /i D ln CCti
i¼1 i¼1 i¼1
X
p
þ /i D ln COCti þ d1 ln Yt1 þ d2 ln PCt1 þ d3 ln CCt1 þ d4 ln COCt1 þ e1t ;
i¼1
(1)

FIG. 2. Coal consumption (million tonnes oil equivalent).

045903-7 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

FIG. 3. Oil consumption (million tonnes).

X
m X
n X
k X
p
D ln PCt ¼ a0 þ bi D ln PCti þ /i D ln Yti þ /i D ln CCti þ /i D ln COCti
i¼1 i¼1 i¼1 i¼1
þ d1 ln PCt1 þ d2 ln Yt1 þ d3 ln CCt1 þ d4 ln COCt1 þ e2t ; (2)

X
k X
m X
n X
p
D ln CCt ¼ a0 þ /i D ln CCti þ bi D ln PCti þ /i D ln Yti þ /i D ln COCti
i¼1 i¼1 i¼1 i¼1

þ d1 ln CCt1 þ d2 ln PCt1 þ d3 ln Yt1 þ d4 ln COCt1 þ e3t ; (3)

X
p X
m X
n X
k
D ln COCt ¼ a0 þ /i D ln COCti þ bi D ln CCti þ /i D ln PCti þ /i D ln Yti
i¼1 i¼1 i¼1 i¼1

þ d1 ln COCt1 þ d2 ln CCt1 þ d2 ln PCt1 þ d3 ln Yt1 þ e4t ; (4)

where D and et are the first difference operator and the white noise term, respectively. An appropri-
ate lag selection is based on the Akaike and/or Schwartz Information Criterion (AIC and SIC). The

FIG. 4. Corbon dioxide emissions (million tonnes).

045903-8 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

bounds testing procedure is based on the joint F-statistic or Wald statistic that tests the null hypoth-
esis of no cointegration. The null hypothesis of no cointegration among the variables in Equations
(1)–(4) is, H0 : d1 ¼ d2 ¼ d3 ¼ d4 ¼ 0, against the alternative hypothesis,
H1 : d1 ¼ 6 d2 ¼6 d3 ¼
6 d4 ¼6 0. When the computed test statistic exceeds the upper critical bounds
value, the H0 hypothesis is rejected. If the F-statistic falls into the bounds, then the cointegration
test becomes inconclusive. If the F-statistic is lower than the lower bounds value, then the null hy-
pothesis of no cointegration cannot be rejected regardless whether the series are I(0) and/or I(1).
As it is accented, we also complement the long-run estimators of the ARDL approach by
two additional long-run estimators by using the FMOLS, DOLS.

2. Granger causality
After estimating the long-run model, the next step is to employ the Granger causality
model, based on the vector error-correction model (VECM). According to Bahmani-Oskooee
et al. (1993), however, if the variables are cointegrated, then the standard Granger causality test
results will be invalid. In this case, the vector error correction model (VECM) should serve as
a starting point for causality analysis.
The VECM used to analyze the relationships between the variables was constructed as
follows:

X
m X
n X
k X
p
D ln Y ¼ a0 þ a1i D ln Yti þ a2i D ln PCti þ a3i D ln CCti þ a4i D ln COCti
i¼1 i¼1 i¼1 i¼1

þ f1i ECMt1 þ et ; (5)

X
m X
k X
p X
n
D ln PC ¼ a0 þ h1i D ln PCti þ h2i D ln CCti þ h3i D ln COCti þ h4i D ln Yti
i¼1 i¼1 i¼1 i¼1

þ f2i ECMt1 þ et ; (6)

X
m X
p X
n X
n
D ln CC ¼ a0 þ b1i D ln CCti þ b2i D ln COCti þ b3i D ln Yti þ b4i D ln PCti
i¼1 i¼1 i¼1 i¼1

þ f3i ECMt1 þ et ; (7)

X
k X
l X
q
D ln COC ¼ a0 þ r1i D ln COCti þ r2i D ln Yti þ r3i D ln PCti
i¼1 i¼1 i¼1
X
p
þ r4i D ln CCti þ f4i ECMt1 þ et ; (8)
i¼1

where residuals et is independently and normally distributed (i.i.d.) with zero mean and constant
variance; ECMt1 is the ECM term resulting from the long-run equilibrium relationship, and f
is a parameter indicating the speed of adjustment to the equilibrium level after a shock. It
shows how quickly variables converge to equilibrium and must have a statistically significant
coefficient with a negative sign.
Granger causality can be examined in short-run Granger causalities by H0 : a2i ¼ a3i
¼ a4i ¼ 0 and H1 : a2i 6¼ a3i 6¼ a4i 6¼ 0 in Equation (5) H0 : h2i ¼ h3i ¼ h4i ¼ 0 and H1 :
h2i 6¼ h3i 6¼ h4i 6¼ 0 in Equation (6), H0 : b2i ¼ b3i ¼ b4i ¼ 0 and H1 : b2i 6¼ b3i 6¼ b4i 6¼ 0
in Equation (7) H0 : r2i ¼ r3i ¼ r4i ¼ 0 and H1 : r2i 6¼ r3i 6¼ r4i 6¼ 0 in Equation (8) for all i.

IV. EMPIRICAL RESULTS

In this paper, the estimation process was constructed in five steps. First, the ARDL bound
test after unit root test was implemented. The unit root tests are used to determine if the
045903-9 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

variables are I(0) and/or I(1). It is important to know the order of the integration of variables
because when the order of integration of any of the variables is known, the critical bounds pro-
vided by Pesaran and Pesaran (1997), Pesaran et al. (2001), and Narayan (2005) can be selected.
In the second step, parameters of the long-run relationship were estimated. In this stage, to
compare results obtained from several tests is very important for our analysis. Two additional
long-run estimators, the dynamic ordinary least squares (DOLS) developed by Stock and
Watson (1993) and the fully modified ordinary least squares (FMOLS) developed by Phillips
and Hansen (1990) methods were added to complement the long-run estimators obtained from
the ARDL. FMOLS and DOLS are single equation methods. FMOLS method tests for “serial
correlation effects” and the “endogeneity” in the regressors resulting from the existence of a
cointegrating relationship. DOLS method tests the problem of second-order asymptotic bias
arising from serial correlation and endogeneity. In the third step, short-run dynamic parameters
were estimated by using an error correction model. In the fourth step, the causality method was
used to specify the direction of causality and in the last step, to test the strength of our causal-
ity test, we used the generalised forecast error variance decomposition analysis developed by
Pesaran and Shin (1998), which is invariant to the ordering of the series.

A. Bounds testing for cointegration

The first step is to study the bound testing results by ARDL methods. Table II shows the
results of the ARDL bounds tests. The null hypothesis of no long term relationship can be
rejected at the 5% level—it was rejected only at the 10% level for some countries—of signifi-
cance when GDP is considered to be the dependent variable and when other variables are the
independent variables.
Lag length supplying the smallest critical value was determined as the lag length of the
model by using the AIC statistic to determine the optimal lag structure for the ARDL model
because of its best performance in small sample (see L€utkepohl (1991)). The lag order must
high enough to reduce the residual serial correlation problems and at the same time, it must be
low enough so that the conditional ECM is not subject to over-parameterisation problems (Ang,
2007; Narayan, 2005; Wolde-Rufael, 2010a; 2010b; and Wolde-Rufael and Menyah, 2010).
Models were determined after applying the Lagrange multiplier test to all possible models.

B. Long-run results: ARDL, FMOLS, and DOLS

Table III displays the long-run elasticities, along with a number of diagnostic tests results
for the underlying ARDL model. The elasticities are interpreted as usual. For Russia, the elas-
ticity of CO2 emission was found to be <1 and was determined as inferior good. For China,
income elasticity was 0.1896. For Brazil, India, and Turkey, the CO2 elasticity of income was
positive and greater than 1 in the long term but was found to be 2.016 for South Africa.
Table III exhibits some diagnostic check results. The Breusch–Godfrey (BG) test deter-
mines that the correlograms of residuals and squared residuals do not indicate any serial corre-
lation problem. Neither the Jarque–Bera (JB) tests nor the ARCH LM test reveals any problems

TABLE II. Bounds testing for cointegration.

Fy (Y;PC,CC,COC) FPC (PC;Y,CC,COC) FCC (CC;PC,Y, COC) FCOC (COC;PC, CC, Y)

Brazil 6.3579 2.1373 1.0549 2.0887

Russia 1.5623 2.2908 0.91209 9.3386
India 6.8877 2.0190 2.0291 1.1358
China 2.5098 1.6708 2.8172 6.4102
Turkey 5.9805 1.8334 2.0118 1.5279
S.Africa 8.7579 1.2940 1.8407 1.8526
045903-10 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

TABLE III. Long-run coefficients for ARDL.

Y PC CC COC R2 J-Ba ARCH test B-Gb RRc

Brazil … 0.300 6.4008 2.377 0.87 1.16 0.3536 21.328 0.42

(3.77) (2.9878) (2.369)
Russia 0.02248 0.7242 0.012023 … 0.85 2.758 0.745 18.15 1.96
(2.178) (3.91) (0.062)
India - 0.8745 1.3450 1.3001 0.78 2.89 0.106 11.33 1.96
(2.09) (2.39) (2.92)
China 0.1896 0.159382 0.911043 … 0.90 3.37 0.201 15.13 2.895
(2.895) (1.968) (21.25)
Turkey … 0.2459 0.19639 1.238 0.93 3.009 0.35 19.85 1.426
(2.86) (2.117) (10.18)
S.Africa … 1.7237 1.9116 2.016 0.89 2.962 0.59 12.69 1.759
(2.77) (5.12) (8.12)

a
J–B test null is normality.
b
B–G test null is no serial correlation.
c
Ramsey RESET.

with the equations. There is no evidence of parameter instability indicated by Ramsey RESET
tests.
The results of ARDL, FMOLS, and DOLS tests are similar in magnitude and sign as can
be seen in Tables III and IV. The results of three cointegration estimators are entirely consistent
with each other; so, we can easily state that the estimated results are robust. As more specific,
the estimated coefficients are statistically significant, having correct signs.

C. Short-run results and error-correction term

ECM coefficients indicate a mechanism to correct the disequilibrium among economic growth,
oil consumption, coal consumption, and CO2 emissions in Table V. They are negative and change
between 0.1259 and 0.50738 to provide stability for the models. The ECM term shows a slow
speed of adjustment to any disequilibrium toward long-run equilibrium in Brazil, Russia, India,
and China.

D. Granger causality results

The ARDL, FMOLS, and DOLS methods do not determine the direction of the causality.
Because of the fact that knowing the direction of the causality between selected variables is
TABLE IV. FMOLS and DOLS results.

FMOLS DOLS

Y PC CC COC Y PC CC COC

Brazil … 0.311 6.474 2.389 … 0.305 6.399 2.310

(3.65) (2.89) (2.54) (2.99) (3.01) (2.47)
Russia 0.02256 0.7158 0.0132 … 0.02248 0.756 0.0144 …
(0.183) (3.12) (0.088) (0.178) (3.22) (0.088)
India … 0.888 1.3778 1.299 … 0.851 1.442 1.375
(2.45) (2.41) (2.87) (2.71) (2.11) (3.13)
China 0.1989 0.1601 0.907 … 0.1874 0.1552 0.9758 …
(2.15) (2.118) (5.12) (2.001) (1.995) (3.985)
Turkey … 0.2554 0.188 1.211 … 0.2444 0.189 1.22
(2.33) (2.23) (8.78) (2.65) (2.71) (9.98)
S.Africa … 1.714 1.954 2.32 … 1.765 1.901 2.125
(2.65) (5.52) (7.25) (2.66) (4.95) (7.89)
045903-11 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

TABLE V. The error-correction representation model.

dY dPC dCC dCOC ECM

Brazil … 0.21346 (2.1889) 0.98732 (2.6643) 1.6471 (5.689) 0.16998 (2.1245)

Russia 0.15767 (2.8974) 0.25656 (4.3725) 0.39211 (7.3541) … 0.1259 (1.8913)
India … 0.44146 (3.2603) 0.35364 (1.7639) 0.49955 (3.4119) 0.20464 (3.967)
China 0.4126  102 (0.188) 0.18274 (3.811) 0.80787 (11.538) … 0.200142 (1.8889)
Turkey 0.12478 (2.1163) 0.099643 (1.9989) 0.89998 (6.6578) 0.50738 (6.7019)

S.Africa 0.13941 (3.366) 0.11786 (1.965) 0.12757 (1.8957) 0.33224 (2.118)

essential for policy implications, we applied Granger causality tests (Table VI). Table VI sum-
marizes the causality relationships between coal and oil consumption, CO2 emission and eco-
nomic growth. Results suggest the evidence of a unidirectional Granger causality from GDP to
CO2 emission in all countries; from coal consumption to CO2 emission in all countries except
India and from oil consumption to CO2 emission in China, India, Turkey, and South Africa and
bi-directional one in Brazil and Russia. Meanwhile, there is bidirectional causality between
GDP and oil consumption and between GDP and coal consumption in all selected countries.
The causality test does not allow to gauge the relative strength of the Granger causality
between the series beyond the sample period (Wolde-Rufael, 2010a; 2010b). For this reason, it
was decomposed the forecast-error variance of economic growth into proportions attributed to
shocks in all variables in the system as suggested by Pesaran and Shin (1998). So the contribu-
tions of oil-coal consumption and CO2 emission to economic growth were determined and the
contribution of economic growth and oil-coal consumption to CO2 emission.
As similar to our results, uni-directional causality from GDP to CO2 emissions is found for
China by Jalil and Mahmud (2009), Chang (2010), and Govindaraju and Tang (2013). Pao and
Tsai (2011) found the evidence of uni-directional causality from GDP to CO2 emissions for
BRIC. For India, Ghosh (2010) and Govindaraju and Tang (2013) found bi-directional causality
between CO2 emissions and economic growth as differentiation to our results. The results
obtained for Brazil, Russia, China, Turkey, and South Africa in causality relationship between
coal consumption and CO2 emissions are similar to Menyah and Wolde-Rufael (2010).

TABLE VI. Results of Granger causality tests for oil consumption and economic growth.

DY ! DCC DY ! DCOC DY ! DPC DCOC ! DCC DCOC ! DPC DPC ! DCC

DCC ! DY DCOC ! DY DPC ! DY DCC ! DCOC DPC ! DCOC DCC ! DPC

Brazil 24.10 15.1344 22.2208 0.239751 2.99435 3.13751

3.010 1.209653 19.5536 11.3098 27.66453 13.0984
Russia 17.11 5.1214 20.768 0.019851 2.4178 19.985
10.22 1.0063 4.2088 9.9987 3.1127 24.5679
India 12.1115 5.6123 41.2398 4.0257 1.5607 0.0777
15.6227 1.913 11.1945 12.345 4.2914 0.1786
China 5.1619 31.1346 50.9123 1.12109 0.9168 7.2570
35.457 0.27834 10.3498 16.983 12.498 21.345
Turkey 7.633 25.667 30.9635 0.01339 1.1108 21.6581
24.96 1.0011 3.51176 19.665 26.3329 32.3409
S.Africa 11.789 9.3451 7.3516 0.1986 1.00985 21.908
10.184 0.01234 5.3398 12.3456 5.1567 14.5678
045903-12 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

E. Generalized forecast error variance decomposition results

We used the generalized forecast error variance decomposition methods suggested by
Pesaran and Shin (1998). The difference between the orthogonalized and generalized forecast
error variance decompositions is that in an orthogonalized forecast error variance decomposi-
tion, the percentage of the forecast error variance of a variable which is accounted for by the
innovation of another variable in the VAR will sum to one across all variables (Wolde-Rufael,
2010a; 2010b and Wolde-Rufael and Menyah, 2010). But the generalized forecast error var-
iance decomposition methods accept one to make robust comparisons of the strength, size, and
persistence of shocks from one equation to another (Payne, 2002). As differentiation from the
orthogonalized case, the row values for the generalized decompositions do not sum up to 100
(Payne, 2002; Sari and Soytas, 2007; and Wolde-Rufael and Menyah, 2010). And this method
provides an “optimal” measure of the amount of forecast error variance decomposition for each
series (see Payne (2002), Sari and Soytas (2007), Wolde-Rufael, 2010a; 2010b, and Wolde-
Rufael and Menyah (2010)).
The results of variance decomposition analysis were given in Table VII. Almost a quarter
of the forecast-error variance of economic growth is explained by oil consumption’s forecast-
error variance as 28.4% for South Africa, 32.6% for China, and 14.68% for India. The forecast
error variance of CO2 emission is explained by more than 10% of the forecast error variance of
GDP in all selected countries. The biggest contribution to the forecast error variance of CO2
emission was made by forecast error variance of GDP (such as in Brazil (37%–42%) and South
Africa (25.2%–29.7%)). These results confirm the unidirectional causality running from GDP to
CO2 emission found by the Granger causality test. The policy implication of our findings is that
policies that reduce CO2 emission do not cause to an adverse impact on economic growth.

V. CONCLUSION
This study has aimed to analyze links between CO2 emissions, economic growth, coal
and oil consumption in BRICTS countries, all of which were reported in the high economic
development. This paper has two contributions. First is the study of the relevant question,
relationship between CO2 emissions, economic growth, and coal and oil consumption in
BRICTS countries. Second is the use of several different methodologies to produce and con-
firm empirical results. The main findings are: (i) The long-run elasticities determined by the
three different methods (ARDL, FMOLS, and DOLS) showed similar results with respect to
the sign and magnificent of the coefficients. (ii) The ECM coefficients indicated a mecha-
nism to correct the disequilibrium among economic growth, oil consumption, coal consump-
tion, and CO2 emissions. ECM coefficients are negative and change between 0.1259 and
0.50738 to provide stability for the model. The ECM term showed a slow speed of adjust-
ment for Brazil, Russia, India, and China. (iii) Since oil and coal consumption plays a vital
role in economic growth both directly and indirectly in all selected countries, energy
conservation-oriented policies that reduce oil and coal consumption can negatively affect
economic growth. These countries need a large amount of oil and coal for sustainable eco-
nomic growth. These countries’ policy makers must determine correctly the energy demand
management policy. (iv) Our results showed that economic growth enhances carbon dioxide
emissions. For BRICTS countries, we found the evidence of uni-directional causality running
from GDP and CO2 emissions in all countries, from coal consumption to CO2 emissions in
all countries except India, and from oil consumption to CO2 emissions in all countries except
Brazil and Russia. Finally, it was obtained the evidence of a uni-directional relation running
from GDP to CO2 emissions for selected countries. In these countries, policies are designed
to reduce CO2 emissions (e.g., efficiency improvement measures) may not have an adverse
impact on economic growth. And the governments can select policies such as cap-trade and
a carbon tax.
Considering the results relating to environmental pollution, three important points stand
out. First, analyzed BRICTS countries must strive to manage environmental pollution. Policy
045903-13 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

TABLE VII. Variance decomposition results.

Dependent variable D lny Dependent variable D lncoc

Horizon Dy D lncc D lnpc D lncoc Dy D lncc D lnpc D lncoc

Brazil
1 98.154 1.8458 1.2530 1.252 36.856 1.81 3.14 82.04
5 97.799 2.6671 2.5336 2.879 41.956 2.52 4.69 78.79
10 82.334 3.1570 16.508 3.562 40.632 2.83 4.43 78.72

Russia
1 80.00 1.526 13.30 5.46 15.68 1.10 2.4614 90.5385
5 77.807 2.856 14.59 6.63 17.26 3.44 3.6299 85.9016
10 69.160 3.458 15.77 9.69 19.12 3.09 8.6888 80.7582

India
1 98.77 4.22 9.69 3.56 13.00 7.859 3.14 96.164
5 71.30 5.02 12.69 4.89 20.589 20.589 4.62 73.282
10 69.36 6.94 14.68 6.02 20.30 30.69 4.37 57.17

China
1 63.250 20.85 12.350 7.250 14.78 11.09 4.181 84.044
5 58.783 19.85 23.87 8.782 16.02 23.30 5.159 76.791
10 57.521 19.56 32.593 9.521 18.96 23.96 6.332 68.727

Turkey
1 75.121 2.1765 2.596 5. 569 28.56 23.77 2.98 90.3125
5 70.716 3.321 5.962 5. 650 20.759 35.125 4.012 88.7695
10 59.779 4.751 7.469 5.965 12.856 39.9001 5.985 88.8971

S.Africa
1 78.5689 9.16 10.61 2.56 25.1968 21.7891 2.856 95.1968
5 55.1153 12.39 20. 61 5.91 29.9001 23.8569 4.02 89.9001
10 51.1964 15. 03 28. 43 7.53 29.6459 23.9985 5.89 89.6459

measures are required to curb environmental pollutants in certain countries. Current energy
policies should be designed from the perspective of sustainable energy. In BRICTS countries,
the appropriate policy must aim to improve the energy infrastructure in the context of the
elasticity and Granger Causality results. In recent years, BRICTS countries have found the
specific environmental organization and/or institutions and made specific laws for different
environmental problems. Second, the change of CO2 emissions may be explained by the rise
of awareness of individuals. Increased education is important in terms of making people aware
of the consequences of environmental pollution. Third, BRICTS countries should cut the sub-
sidies on oil and adopt “the true prices” policy to lead consumers to the less-pollutant and
low-cost energy sources. Economic growth must be compatible with improvements in environ-
mental quality.
At the same time, our results show that environmental problems in these countries will
inevitability impact other fossil fuel industries. Natural gas in fossil energy, for example, poses
advantages in terms of efficiency and CO2 emissions vis-a-vis sustainable energy.

Akbostanci, E., Turut-Asik, S., and Tunc, G. I., “The relationship between income and environment in Turkey: Is there an
environmental Kuznets curve,” Energy Policy 37, 861–867 (2009).
Alam, M. J., Begum, I. A., Buysse, J., Rahman, S., and Van Huylenbroeck, G., “Dynamic modeling of the causal relation-
ship between energy consumption, CO2 emissions and economic growth in India,” Renewable Sustainable Energy Rev.
15, 3243–3251 (2011).
Ang, J. B., “CO2 emissions, energy consumption, and output in France,” Energy Policy 35, 4772–4778 (2007).
045903-14 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

Ang, J. B., “Economic development, pollutant emissions and energy consumption in Malaysia,” J. Policy Model. 30,
271–278 (2008).
Apergis, N. and Payne, J. E., “The emissions, energy consumption, and growth nexus; evidence from the commonwealth of
independent states,” Energy Policy 38, 650–655 (2010).
Arouri, M., Lahiani, A., and Nguyen, D. K., “Return and volatility transmission between world oil prices and stock markets
of the GCC countries,” Econ. Modell. 28, 1815–1825 (2011).
Bahmani-Oskooee, M., Mohtadi, H., and Alse, J., “Export growth and economic growth: An application of cointegration
and error-correction modeling,” J. Dev. Areas 27, 535–542 (1993).
Bildirici, M. E. and Bakirtas, T., “The relationship among oil, natural gas and coal consumption and economic growth in
BRICTS (Brazil, Russian, India, China, Turkey and South Africa) countries,” Energy 65, 134–144 (2014).
Bloch, H., Rafiq, S., and Salim, R., “Coal consumption, CO2 emission and economic growth in China: Empirical evidence
and policy responses,” Energy Econ. 34, 518–528 (2012).
BP, see http://www.bp.com/content/dam/bp-country/de_de/PDFs/brochures/BP-statistical-review-of-world-energy-2014-
full-report.pdf for BP Statistical Review of World Energy June 2014 (2014).
Chang, C. C., “A multivariate causality test of carbon di oxide emission, energy consumption and economic growth in
China,” Appl. Energy 87, 3533–3537 (2010).
Cole, M. A., Rayner, A. J., and Bates, J. M., “The environmental Kuznets curve: An empirical analysis,” Environ. Dev.
Econ. 2, 401–416 (1997).
Connors, S. R., “Issues in energy and sustainable development,” presented at the Fourth International Programme on the
Management of Sustainability, Netherlands, June 1977, Report No. EL 98-004, 1998.
De Bruyn, S. M., van den Bergh, J. C. J. M., and Opschoor, J. B., “Economic growth and emissions: Reconsidering the em-
pirical basis of environmental Kuznets curves,” Ecol. Econ. 25, 161–175 (1998).
Emran, M. S., Shilpi, F., and Alam, M. I., “Economic liberalization and price response of aggregate private investment:
Time series evidence from India,” Can. J. Econ. 40, 914–934 (2007).
Fodha, M. and Zaghdoud, O., “Economic growth and pollution emissions in Tunisia: An empirical analysis of the environ-
mental Kuznets curve,” Energy Policy 38, 1150–1156 (2010).
Ghosh, S., “Examining carbon emissions economic growth nexus for India: A multivariate cointegration approach,”
Energy Policy 38, 3008–3014 (2010).
Govindaraju, V. G. R. C. and Tang, C. F., “The dynamic links between CO2 emissions, economic growth and coal con-
sumption in China and India,” Appl. Energy 104, 310–318 (2013).
Grossman, G. M. and Krueger, A. B., “Environmental impacts of a North American free trade agreement,” in National
Bureau of Economic Research Working Paper 3914, NBER, Cambridge, MA (1991).
Halicioglu, F., “An econometric study of CO2 emissions, energy consumption, income and foreign trade in Turkey,”
Energy Policy 37(3), 1156–1164 (2009).
Jalil, A. and Mahmud, S. F., “Environment Kuznets curve for CO2 emissions: A cointegration analysis for China,” Energy
Policy 37, 5167–5172 (2009).
Jinke, L., Hualing, S., and Dianming, G., “Causality relationship between coal consumption and GDP: Difference of major
OECD and non-OECD countries,” Appl. Energy 85, 421–429 (2008).
Kulshreshtha, M. and Parikh, J. K., “Modeling demand for coal in India: Vector autoregressive models with cointegrated
variables,” Energy 25, 149–168 (2000).
Li, R. and Leung, G. C. K., “Coal consumption and economic growth in China,” Energy Policy 40, 438–443 (2012).
L€utkepohl, H., Introduction to Multiple Time Series Analysis (Springer-Verlag, 1991).
Menyah, K. and Wolde-Rufael, Y., “Energy consumption, pollutant emissions and economic growth,” Energy Econ. 32(6),
1374–1382 (2010).
Narayan, P. K., “The saving and investment nexus for China: Evidence from cointegration tests,” Appl. Econ. 37,
1979–1990 (2005).
Pao, H.-T. and Tsai, C.-M., “Multivariate Granger causality between CO2 emissions, energy consumption, FDI (foreign
direct investment) and GDP (gross domestic product): Evidence from a panel of BRIC (Brazil, Russian Federation, India,
and China) countries,” Energy 36, 685–693 (2011).
Paul, S. and Bhattacharya, R. N., “Causality between energy consumption and economic growth in India: A note on con-
flicting results,” Energy Econ. 26, 977–983 (2004).
Payne, J. E., “Inflationary dynamics of a transition economy: The Croatian experience,” J. Policy Model. 24, 219–230
(2002).
Pesaran, H. M. and Pesaran, B., Microfit 4.0 (Oxford University Press, Oxford, 1997).
Pesaran, M. H. and Shin, Y., “Generalised impulse response analysis in linear multivariate models,” Econ. Lett. 58, 17–29
(1998).
Pesaran, M. H., Shin, Y., and Smith, S., “Bounds testing approach to the analysis of level relationships,” J. Appl. Econ. 16,
289–326 (2001).
Phillips, P. C. B. and Hansen, B. E., “Statistical inference in instrumental variables regression with I(1) processes,” Rev.
Econ. Studies 57, 99–125 (1990).
Richmond, A. K. and Kaufmann, R. K., “Is there a turning point in the relationship between income and energy use and/or
carbon emissions?,” Ecol. Econ. 56, 176–189 (2006).
Saikkonen, P., “Asymptotically efficient estimation of cointegration regressions,” Econometric Theory 7, 1–21 (1991).
Sanjari, F. and Delangizan, S., see http://ssrn.com/abstract¼1635233 for “Carbon Emissions and Economic Growth: The
Iranian Experience (2010).”
Sari, R. and Soytas, U., “The growth of income and energy consumption in six developing countries,” Energy Policy 35,
889–898 (2007).
Selden, T. M. and Song, D., “Environmental quality and development: Is there a Kuznets curve for air pollution?,”
J. Environ. Econ. Manage. 27, 147–162 (1994).
Shafik, N., “Economic development and environmental quality: An econometric analysis,” Oxford Econ. Pap. 46, 757–773
(1994).
045903-15 E. M. Bildirici and T. Bakirtas J. Renewable Sustainable Energy 8, 045903 (2016)

Shafik, N. and Bandyopadhyay, S., “Economic growth and environmental quality: Time series and cross-country evidence,”
Background Paper for the World Development Report 1992, Report No. WPS904, The World Bank, Washington, DC,
1992.
Shahbaz, M., Lean, H. H., and Shabbir, M. S., “Environmental Kuznets curve and the role of energy consumption in
Pakistan,” in Development Research Unit Discussion Paper DEVDP 10/05 (2010).
Soytas, U. and Sari, R., “Energy consumption, economic growth, and carbon emissions: Challenges faced by an EU candi-
date member,” Ecol. Econ. 68, 1667–1675 (2009).
Soytas, U., Sari, R., and Ewing, B. T., “Energy consumption, income, and carbon emissions in the United States,” Ecol.
Econ. 62, 482–489 (2007).
Stock, J. H. and Watson, M. W., “A simple estimator of cointegrating vectors in higher order integrated systems,”
Econometrica 61, 783–820 (1993).
Tucker, M., “Carbon dioxide emissions and global GDP,” Ecol. Econ. 15, 215–223 (1995).
Wang, S. S., Zhou, D. Q., Zhou, P., and Wang, Q. W., “CO2 emissions, energy consumption and economic growth in
China: A panel data analysis,” Energy Policy 39, 4870–4875 (2011).
Wolde-Rufael, Y., “Coal consumption and economic growth revisited,” Appl. Energy 87, 160–167 (2010a).
Wolde-Rufael, Y., “Bounds test approach to cointegration and causality between nuclear energy consumption and eco-
nomic growth in India,” Energy Policy 38, 52–58 (2010b).
Wolde-Rufael, Y. and Menyah, K., “Nuclear energy consumption and economic growth in nine developed countries,”
Energy Econ. 32, 550–556 (2010).
Yuan, J.-H., Kang, J.-G., Zhao, C.-H., and Hu, Z.-G., “Energy consumption and economic growth: Evidence from China at
both aggregated and disaggregated levels,” Energy Econ. 30, 3077–3094 (2008).
Zhang, X.-P. and Cheng, X.-M., “Energy consumption, carbon emissions, and economic growth in China,” Ecol. Econ. 68,
2706–2712 (2009).