Energy 39 (2012) 180e186

Contents lists available at SciVerse ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

The impact of energy consumption and CO2 emission on the economic growth
and financial development in the Sub Saharan African countries
Usama Al-mulali a, *, Che Normee Binti Che Sab b
a
Economic Programe, Universiti Sains Malaysia, 11800 USM, Malaysia
b
Economic Section, Universiti Sains Malaysia, 11800 USM, Malaysia

a r t i c l e i n f o a b s t r a c t

Article history: This study investigated the impact of energy consumption and CO2 emission on GDP (gross domestic
Received 16 August 2011 product) growth and the financial development in thirty Sub Saharan African Countries. The panel model
Received in revised form was used in this study from the period 1980 to 2008. The results showed that energy consumption had
13 January 2012
played an important role to increase both economic growth and the financial development in the
Accepted 17 January 2012
Available online 16 February 2012
investigated economies but with the consequence of high po llution. This study recommended that these
countries should increase energy productivity by increasing energy efficiency, implementation of energy
savings projects, energy conservation, and energy infrastructure outsourcing to achieve its financial
JEL classifications:
O1
development and GDP growth and to increase their investment on energy projects to achieve the full
Q43 energy potential.
Q53 Ó 2012 Elsevier Ltd. All rights reserved.
Q56
Keywords:
Total primary energy consumption
CO2 emission
GDP growth
Financial development
Sub Saharan African countries

1. Introduction Malawi, Mali, Mauritius, Niger, Nigeria, Rwanda, Senegal, Sierra

Leone, South Africa, Swaziland, Togo, and Zambia.
Sub Saharan African (SSA) contains countries that have the
lowest gross domestic product in the world as well as the lowest
2. Studies on energy consumption and GDP growth
economic and financial development. However, based on the World
Economic Development (WED), the countries of this reign achieved
The relationship between the energy consumption and GDP
improvements in growth in the 2000’s, managed to achieve high
growth was the focus of research conducted by many researchers.
GDP (gross domestic product) growth by 5.08% in 2008, and the
Warr and Ayres [1] found one way causal relationship from energy
GDP per capita rose by 2.5%. Moreover, the energy consumption of
consumption to economic growth in the US, in addition, no
this reign rose by 6% and CO2 emission increased by 20%. This point
evidence of causal relationship from economic growth to energy
helps us to reach the main goal of this study which is to examine
consumption was found. Therefore, to sustain long run growth, it is
the impact of energy consumption on the economic growth and the
important to increase either energy suppliers or the efficiency of
financial development in the Sub Saharan African countries,
energy usage. While Fallahi [2] found a strong bi-directional causal
namely, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape
relationship between energy consumption and GDP growth in the
Verde, Central African Republic, Chad, Comoros, Republic of Congo,
United States. Also one way causal relationship is present from
Ethiopia, Gabon, Gambia, Ghana, Kenya, Lesotho, Madagascar,
biomass energy consumption to economic growth in the United
States [3]. Bownden and Payne [4] found one directional causal
relationship from non-renewable industrial energy consumption to
* Corresponding author. Tel.: þ60 174587786.
GDP growth in the US. On the contrary, bi-directional causal rela-
E-mail addresses: [email protected] (U. Al-mulali), [email protected] tionship was found between oil consumption and economic growth
(C.N. Binti Che Sab). in Korea [5]. Lee and Chang [6] found a strong positive relationship

0360-5442/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.energy.2012.01.032
 U. Al-mulali, C.N. Binti Che Sab / Energy 39 (2012) 180e186 181

between energy consumption and economic growth in Taiwan. You economic expansion is the biggest contributor to the rise in CO2
[7] found that the renewable energy consumption has a strong emission in Turkey. The carbon intensity and energy consumption
relationship in achieving sustainable development in China for the also contribute to the rise of the CO2 emission. Freitas and Kaneko
economic societies by its positive effect on the country’s genuine [22] found similar results in Brazil. However, it was also found that
savings and economic growth whereas the increase in fossil fuel the increase in the consumption of clean energy reduces CO2
energy consumption increases China’s economic growth only. In emission. Niu et al. [23] found a long run relationship between
the Sub Saharan African countries it was found that energy energy consumption, CO2 emission and economic growth in eight
consumption has inverse relationship with the industrial devel- Asian economies, however, despite the fact that CO2 emission per
opment, but a positive relationship to GDP [8]. Yingzi and Yuying capita and energy efficiency of energy use in the developing
[9] found that China high economic development increased both countries is much lower than the developed countries, the CO2
energy consumption and CO2 emission in the last three decades. emission per unit of energy use is much higher than the developed
The same results were found by Yuan et al. [10]. Table 1 below countries. In Iran, a one way causal relationship from economic
shows the results arrived at by early studies dealing with the growth and energy consumption (petroleum products and natural
relation between energy consumption, GDP growth. gas consumption) to CO2 emission was found. However, there was
no causal relationship running from fossil fuels consumption to CO2
3. Studies on energy consumption, CO2 emission, and GDP emission. Moreover, no evidence that CO2 emission, petroleum
growth products, fossil fuel consumption led to economic growth [24]. In
Malaysia, it was found that the energy consumption and pollution
A large number of studies found the existence of a causal rela- increase economic output, in addition it was found that causal
tionship between energy consumption, CO2 emission and economic relationship is present running from economic growth to energy
growth such as China by Chang [11]. Similar results were found in consumption in the short run and the long run [25]. Table 2 shows
Turkey by Halicioglu [12] who also found that the income had a summary of what has been previously mentioned regarding the
a more significant impact in explaining the CO2 emission in Turkey results arrived at by the studies that deal with energy consumption,
than the energy consumption. Pao and Tsai [13] found similar CO2 emission and GDP Growth.
results in Brazil. Lean and Smyth [14] found a causal relationship
running from electricity consumption and CO2 emission to 4. Studies on energy consumption, CO2, and the financial
economic output, and causal relationship exists between CO2 development
emission to energy consumption in the ASEAN countries. Ozturk
and Acaravci [15] found similar results in Turkey where a short run Another number of studies examined the relationship between
and long run causal relationship between energy consumption, CO2 energy consumption and the financial development which plays an
emission and growth exists. Similar results were found in the BRIC important role in achieving high economic growth Shumpeter [26],
countries by Pao and Tsai [16]. While one directional causal rela- Goldsmith [27], McKinnon [28] and Shaw [29]. Jalil and Feridun
tionship from GDP to energy consumption and from energy [30] found that the increase in the financial development decreases
consumption to CO2 emission exists in China [17]. Menyah and environmental pollution. Similar results were found by Tamazian
Rufael [18] found a long run and a one directional casual relation- et al. [31] in the BRIC countries, however, Zhang [32] found that the
ship from energy consumption and CO2 emission to economic financial development increases CO2 emission in China. Sadorsky
growth in South Africa. Pao et al. [19] also found similar results in [33] found that the financial development increased energy
Russia. Menyah and Rufael [20] suggested that nuclear energy consumption in Central and Eastern Europe countries by using
consumption can reduce CO2 emission. Lise [21] found that the stock market and banking variables as indicators of the financial

Table 1
Summary of results arrived at by related earlier studies on energy consumption, and GDP growth.

Authors Country Period Variables Methodology Conclusion

Warr & Ayres USA 1946e2000 GDP, capital, labour, energy Granger causality VEC, JJ cointegration EC / GDP
consumption GDP no/ EC
Fallahi USA 1960e2005 GDP, energy consumption Granger causality MS-VAR model GDP 4 EC
Payne USA 1949e2007 Biomass energy consumption, GDP TodaeYamamoto Granger causality BEC / GDP
Bownden & Payne USA 1949e2006 GDP, renewable and TodaeYamamoto Granger causality REC 4 GDP
non-renewable energy NREC 4 GDP
consumption, gross fixed
capital formation, employment
Yoo South Korea 1968e2002 Oil consumption, GDP Granger causality VEC GDP 4 OCON
Lee & Chang Taiwan 1953e2003 GDP, capital stock, labour, Liner model EC have positive effect on GDP
export, energy consumption
You China 1980e2003 GDP per capita, fossil fuel Analysis of variance decomposition, FCON increase GDPP, REC increases GS
consumption, renewable generalized impulse response function
energy consumption,
genuine saving
Kebede et al. SSA 1980e2004 GDP, population, energy The Parks model EC have positive effect on GDP & GDPP
consumption, oil price,
industrial share, agriculture share,
Yingzi & Yuying China 1985e2009 GDP, energy consumption, OLS EC have positive effect on GDP & IWW
industrial waste water
Yuan et al. China 1985e2007 GDP, energy consumption Grey incidence analysis EC have positive effect on GDP

Note: / is one directional causal relationship, lnp is the long run positive relationship, CO2 is carbon dioxide emission FCON is the fossil fuel consumption, JJ is the
JohanseneJuselius, REC is the renewable energy consumption, GS is the genuine saving, IWW is the industrial water waste, SSA is the Sub Saharan African countries. VAR is
vector autoregressive model, VEC is the vector error correction model, TY is TodaeYamamoto, ARDL is autoregressive distributed lag, GDPP is the gross domestic product per
capita, EC is energy consumption, CO2 is carbon dioxide emission, FT is foreign trade, BEC is the biomass consumption, REC is the renewable energy consumption, NREC is the
non-renewable energy consumption, FCON is the fossil fuel consumption, OCN is the oil consumption.
182 U. Al-mulali, C.N. Binti Che Sab / Energy 39 (2012) 180e186

Table 2
Summary of results arrived at by related earlier studies on energy consumption, CO2 emission and growth.

Authors Country Period Variables Methodology Conclusion

Chang China 1982e2004 GDP, energy consumption, Multivariate cointegration, GDP 4 CO2
CO2 emission vector error correction EC 4 GDP
model EC / CO2
Halicioglu Turkey 1960e2005 CO2 emission, energy Granger causality ARDL CO2 4 GDP
consumption, per capita cointegration EC s GDP
income, foreign trade FT s CO2
Lean & Smyth ASEAN 1980e2006 CO2 emission, electricity Johansen Fisher panel ELC lnp CO2
consumption, GDP per capita cointegration, Panel VEC ELC / GDP
Granger causality CO2 / ELC
Ozturk & Acaravci Turkey 1968e2005 GDP, CO2 emission, energy Granger causality ARDL EC s GDP
consumption, employment cointegration CO2 s GDP
EM / GDP
Pao & Tsai BRIC 1971e2005 GDP, CO2 emission, energy Panel Pedroni cointegration, EC lnp CO2
consumption Panel VEC Granger causality EC 4 CO2
EC 4 GDP
CO2 / GDP
Zhang & Cheng China 1960e2007 GDP, gross fixed capital Granger causality; VEC; GDP / EC
formation, energy generalized EC / CO2
consumption, CO2 emission, impulse response, VAR EC no/ GDP
urban population CO2 no/ GDP
Menyah & Rufael South Africa 1965e2006 GDP, CO2 emission, energy Granger causality ARDL CO2 lnp GDP
consumption, gross fixed cointegration CO2 / GDP
capital formation, EC / GDP
employment EC / CO2
Pao & Tsai Brazil 1980e2007 CO2 emission, energy Cointegration Grey, Granger GDP lnp CO2
consumption, GDP causality, prediction EC lnp CO2
model (GM) GDP 4 CO2
GDP4 EC
EC 4 CO2
Pao et al. Russia 1990e2007 GDP, energy consumption, Granger causality VEC, GDP 4 CO2
CO2 emission JJ cointegration GDP 4 EC
EC 4 CO2
Menyah & Rufael USA 1960e2007 CO2 emission, renewable Modified Granger EC / CO2
energy consumption, causality VEC REC s CO2
nuclear energy NEC / CO2
consumption, GDP
Lise Turkey 1980e2003 GNP, CO2 emissions OLS CO2 have positive effect on GDP
Freitas & Kaneko Brazil 1970e2009 CO2 emission, GDP, Decomposition approach GDP and POP have positive effect CO2
population and model
formulation
Niu et al. APC 1971e2005 GDP, energy consumption, Panel Pedroni cointegration, EC 4 GDP
CO2 emission Panel VEC Granger causality EC / CO2
GDP / CO2
Lotfalipour Iran 1967e2007 CO2 emission, fossil fuel Granger causality VAR GDP / CO2
consumption, GDP FCON / CO2
Ang Malaysia 1971e1999 GDP per capita, energy VEC Granger causality, CO2 & EC lnp GDPP
consumption, JJ cointegration GDPP / EC
CO2 emission

Note: 4, /, s, no/, lnp is the bi-directional causal relationship, one directional causal relationship, no causal relationship, no unidirectional causal relationship, and long
run positive relationship. VAR is vector autoregressive model, VEC is the vector error correction model, JJ is the JohanseneJuselius, TY is TodaeYamamoto, ARDL is autore-
gressive distributed lag, EC is energy consumption, CO2 is carbon dioxide emission, FT is foreign trade, EM is the employment ratio, ECL is the electricity consumption, BEC is
the biomass consumption, REC is the renewable energy consumption, NEC is the nuclear energy consumption, OCN is the oil consumption, APC is the Asian pacific countries.

development. The same researcher in a different study found the behind building the GDP per capita model is to examine the impact
same results in twenty-two emerging economies [34]. of total energy consumption impact on economic growth. Since the
Table 3 shows a summary of the results arrived at by early previous studies had only focused on the impact of the financial
studies that deal with energy consumption, CO2 emission and the development on CO2 emission, the present study tries to explore
financial development. how the domestic credit can be related to the private sector and the
broad money models. The purpose behind this is to examine the
impact of total primary energy consumption and CO2 emission on
5. Methodology the financial development in the SSA countries. This study also
attempts to examine the impact of the GDP growth and the
The panel model will be implemented taking the period financial development on total energy consumption and CO2
1980e2008. Three models will be built in this study; the GDP per emission. The models are presented as follows:
capita model as an indicator of economic growth, the broad money
model and private credit to the private sector as indicators of the GDPPit ¼ a þ b1 ECit þ b2 EMit þ b3 INVit þ εit (1)
financial development. During the period 1980e2008, the coun-
tries under investigation had managed to increase their financial BMit ¼ a þ b1 ECit þ b2 EMit þ b3 INVit þ εit (2)
development, GDP growth, total energy consumption, and CO2
emission (World Development Indicators). Thus, the motivation DCPSit ¼ a þ b1 ECit þ b2 EMit þ b3 INVit þ εit (3)
 U. Al-mulali, C.N. Binti Che Sab / Energy 39 (2012) 180e186 183

Table 3
Summary of results arrived at by related earlier studies on energy consumption, CO2 emission and financial development.

Authors Country Period Variables Methodology Conclusion

Jalil & Feridun China 1953e2006 CO2 emission, energy consumption, Granger causality ARDL GDP / CO2 FDI have negative
GDP, ratio of liquid liabilities from cointegration effect on CO2
GDP, trade ratio, the private sector
loans to the nominal GDP
Tamazian et al. BRIC 1992e2004 CO2 emission, GDPP, gross domestic The standard reduced-form FDI & EDI have negative
expenditure in research and modeling approach form. effect on CO2
development as percentage of GDP
in country, indicates stock market
value added, foreign direct investment,
the ratio of deposit money bank assets
to GDP the capital account
convertibility, the financial
liberalization financial openness,
oil consumption, the energy
consumption
Zhang China 1980e2005 CO2 emission, GDP per capita, the ratio VEC Granger causality, GDP & FDI lnp CO2
of loans in financial intermediation JJ cointegration FDI / CO2
to GDP, the ratio of the sum of loans
to township enterprises, enterprises
with foreign funds and private
enterprises and self-employed
individuals to GDP, the ratio of stock
market capitalization to the GDP,
the ratio of stock market turnover to
GDP, foreign direct investment
inflows as percent of GDP
Sadorsky CEFC 1996e2006 GDP per capita, energy consumption, Generalized method of FDI have positive effect on EC
oil price, deposit money bank assets moments (GMM) regression
to GDP, financial system deposits to
GDP, liquid liabilities to GDP, private
credit by deposit money banks to GDP,
stock market cap to GDP, stock market
value traded to GDP, stock market turnover.
Sadorsky EME 1990e2009 Energy consumption, GDPP, inflation Generalized method of FDI have positive effect on EC
rate, bank deposits to GDP, stock market moments (GMM) regression
cap to GDP, stock market value traded
to GDP, stock market turnover to GDP

Note: / is one directional causal relationship, lnp is the long run positive relationship, FDI is the financial development indicators, CO2 is carbon dioxide emission, CEFC is the
Central and Eastern European Countries, EME is the Emerging Market Economies, VEC is the vector error correction model, JJ is the JohanseneJuselius.

GDPP is the gross domestic product per capita measured in 5.1. Panel unit root test
constant national currency. BM is the broad money measured in
billions of constant national currency. DCPS is the domestic credit Though the unit root test is the most important used test among
to the private sector measured in billions of constant national economist and econometricians, the panel unit root testing is more
currency. EC is the total primary energy consumption measured in recent. Panel unit root testing has become one of the most popular
Quadrillion of Btu. EM is the carbon dioxide emissions from the tests among researchers because it is more powerful than the
consumption of energy measured in million metric tons. INV the normal time series unit root. Two types of unit root tests, namely, the
domestic investment measured in billions of constant national Im, Pesaran, and Shin, and the ADF e Fisher Chi-square will be used.
currency. The b1, b2 and b3 are the coefficient vectors and a is These tests are used to derive a panel specific result. These unit root
a scalar. The data about GDPP and INV, BM and DCPS, and the EM tests specify a separate ADF regression for each cross section:
and EC of Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape Xpi
Verde, Central African Republic, Chad, Comoros, Republic of Congo, Dyit ¼ ayit1 b Dyitj
j¼1 ij
þ Xit0 d þ 3 it (4)
Ethiopia, Gabon, Gambia, Ghana, Kenya, Lesotho, Madagascar,
Malawi, Mali, Mauritius, Niger, Nigeria, Rwanda, Senegal, Sierra The null hypothesis can be written as follows:
Leone, South Africa, Swaziland, Togo, and Zambia are taken from H0: a ¼ 0, for all i.
World Economic Outlook (WEO), the World Development Indica- The alternative hypothesis can be written as follows:
tors (WDE), and the Energy Information Administration (EIA) 
ai ¼ 0 for i ¼ 1; 2; 3..; N1
successively. H1 :
ai < 0 for i ¼ N þ 1; N þ 2; ::; N
The unit root test will be tested to examine whether the vari-
ables contain a panel unit root. If the variables contain a unit root, where the i may be reordered as necessary which may be inter-
the cointegration test will be used to examine whether the long run preted as a non-zero fraction of the individual processes.
relationship is present between the variables. If the long run
relationship is present, the vector error correction Granger 5.2. Panel cointegration test
causality will be tested to find out the causal relationship between
the variables. The explanation of the tests below is taken from If the variables contain a panel unit root, the cointegration test is
Baltagi [35]. used. The Pedroni test is employed in this study to examine
184 U. Al-mulali, C.N. Binti Che Sab / Energy 39 (2012) 180e186

Table 4 allow for heterogenous intercepts and trend coefficients across

Panel unit root test. cross-sections. Consider the following regression:
Variable Level First difference
yit ¼ ai þ dit þ b1 x1;it þ b2 x2;it þ / þ bki xk;it þ 3 it (5)
Intercept Intercept Intercept Intercept
and trend and trend
where t ¼ 1, ., T; i ¼ 1, ., N; j ¼ 1, ., k; and y and x are assumed to
Im, Pesaran and Shin W-stat be integrated of order 1 i.e. I(1). The parameters ai and di are
GDPP 4.01162 1.82315 11.2641*** 11.0321***
individual entity and time effects respectively which may be set to
BM 10.5113 5.92530 7.50677*** 7.22997***
DCPS 2.47867 1.44559 11.3066*** 9.84868*** zero if desired. Under the null hypothesis of no cointegration, the
EC 7.29128 1.33211 10.5080*** 9.61559*** residuals εit will be I(1). Pedroni constructs various statistics for
EM 6.08753 1.36476 10.8580*** 10.1998*** testing the null hypothesis of no cointegration i.e. H0: ri ¼ 1 for all i,
INV 8.06844 5.24461 6.64848*** 6.09968***
where ri is the coefficient of the estimated residual. There are two
ADF e Fisher Chi-square different alternative hypotheses. The first one states that the
GDPP 51.4336 61.2521 245.820*** 231.853*** cointegrating vector bi is homogenous i.e. H1: ri ¼ r < 1 for all i
BM 14.0857 28.8875 183.313*** 175.338***
which Pedroni terms the within-dimension or panel statistics test.
DCPS 51.7321 49.3891 253.087*** 215.283***
EC 30.2988 61.6102 229.339*** 193.685*** The second one states that the cointegrating vector bi is heteroge-
EM 50.5684 55.2149 234.513*** 203.830*** nous i.e. H1: ri < 1 for all i which Pedroni terms the between-
INV 30.0608 61.6921 170.386*** 142.910*** dimension or group statistics test. These cointegration statistics
Note: *** denotes statistical significance at 1% level, we used the Schwarz automatic are The Panel v-statistics, Panel p-statistics, Panel t-statistic (non-
selection of the lag length for the unit root test. parametric), Panel t-statistic (parametric), Group r-statistics, Group
t-statistics (non-parametric), and the Group statistics (parametric).
The first four cointegration statistics refer to the within-dimension
or panel statistics test and the last three refer to the between-
whether the total primary energy consumption and CO2 emission
dimension or group statistics test.
have a long run relationship with the GDP growth and financial
development in the selected SSA countries. A number of studies
which are previously mentioned find a long run positive relation- 5.3. Panel Granger causality
ship between total primary energy consumption, CO2 emission, and
GDP growth. Furthermore, the same test is used to examine The Granger causality in general is to find out whether a variable
whether a long run relationship between the total primary energy x causes a variable y, to see how much of the current value of y can
consumption, CO2 emission, and the financial development in the be explained by past values of y and whether adding past values of x
investigated countries exists. can improve the explanation of y. The variable x is said to Granger-
The panel cointegration is considered much better than the time cause variable y if past values of x help in the prediction of the
series cointegration because it shows the long run relationship present value of y.
between the variables for N (2) countries. The Pedroni The panel Granger causality is used to examine the positive or
(EngleeGranger based) cointegration tests will be used in this study the negative causal relationship between the total primary energy
due to the fact that Pedroni made several tests for cointegration that consumption, CO2 emission, and GDP growth in the selected SSA

Table 5
Pedroni cointegration test results.

Statistic Prob. Weighted statistic Prob.

GDPP Model Alternative hypothesis: common AR coefs. (within-dimension)
Panel v-statistic 2.773293*** 0.0085 1.992640** 0.0548
Panel rho-statistic 0.659667 0.3209 0.718513 0.3082
Panel PP-statistic 5.516350*** 0.0000 3.760501*** 0.0003
Panel ADF-statistic 5.750197*** 0.0000 4.715343*** 0.0000
Alternative hypothesis: individual AR coefs. (between-dimension)
Group rho-statistic 2.594993** 0.0138
Group PP-statistic 3.381022** 0.0013
Group ADF-statistic 4.605800*** 0.0000
BM Model Alternative hypothesis: common AR coefs. (within-dimension)
Panel v-statistic 5.568633*** 0.0000 3.006093** 0.0044
Panel rho-statistic 3.867839*** 0.0002 4.019189*** 0.0001
Panel PP-statistic 2.120249** 0.0421 0.601763 0.3329
Panel ADF-statistic 1.078025 0.2231 0.540538 0.3447
Alternative hypothesis: individual AR coefs. (between-dimension)
Group rho-statistic 5.649981*** 0.0000
Group PP-statistic 1.237950* 0.1854
Group ADF-statistic 0.977224 0.2475
DCPS Model Alternative hypothesis: common AR coefs. (within-dimension)
Panel v-statistic 4.593456*** 0.0000 1.801051** 0.0788
Panel rho-statistic 5.629842*** 0.0000 4.430266*** 0.0000
Panel PP-statistic 4.953502*** 0.0000 0.015299 0.3989
Panel ADF-statistic 11.44001*** 0.0000 2.337651** 0.0260
Alternative hypothesis: individual AR coefs. (between-dimension)
Group rho-statistic 6.020971*** 0.0000
Group PP-statistic 0.025275 0.3988
Group ADF-statistic 1.403128* 0.1491

***, **, * denote significance at 1%, 5% and 10%, we use the automatic selection based on the Schwarz to choose the optimal lag length.
 U. Al-mulali, C.N. Binti Che Sab / Energy 39 (2012) 180e186 185

Table 6
Panel Granger causality test results.

The independent variables

DGDPP DBM DDCPS DEC DEM DINV ect (1)

The dependent variables DGDPP e 3.594962*** 5.395464*** 0.589447 0.666792 11.97863*** 4.500108***
DBM 0.348583 e 0.6385329 2.281958** 2.633004** 4.854702*** 3.356877***
DDCPS 0.295843 0.682394 e 3.635071*** 3.213875*** 4.878198*** 6.071295***
DEC 0.642471 17.84238*** 3.714543*** e 482.1126*** 3.908658*** 1.721424**
DEM 0.518045 10.483663*** 4.748480*** 514.0639*** e 3.068935*** 2.190765**
DINV 8.854229*** 8.394862*** 2.737563*** 2.475372*** 2.519680*** e 3.666473***

Note: The null hypothesis is that there is no causal relationship between variables. Values in parentheses are p-values for Wald tests with a c2 distribution. D is the first
difference operator. ect (1) represents the error correction term lagged one period. *** indicates 1% and ** indicates 5% significant level, D is the first difference operator.

countries. While a large number of studies found a bi-directional 6.2. Panel Granger causality test results
causal relationship between the total primary energy consump-
tion, CO2 emission and GDP growth in other countries, only few Table 6 below reviews the Granger causality test results which
studies examined the causal relationship between the total primary show the existence of a bi-directional causal relationship between
energy consumption, CO2 emission, and the financial development. all the variables in the long run based on the error correction term.
Thus, this study aims to use this test to examine the causal rela- The results also show that in the short run there is a positive bi-
tionship between the total primary energy consumption, CO2 directional causal relationship between total primary energy
emission, and the financial development in the investigated consumption and CO2 emission, GDP per capita and investment,
countries. The Granger causality is commonly used to examine the broad money and total primary energy consumption, CO2 emission
direction of the Granger causality of the variables in the panel and broad money, investment and broad money, domestic credit to
model. Engle and Granger showed that non-stationary variables are the private sector and total primary energy consumption, CO2
cointegrated in the model; hence, the vector error correction model emission and domestic credit to the private sector, investment and
will be used to investigate the temporal short run causality domestic credit to the private sector. The results also show that
between the variables. Short run Granger causality can be estab- there is a one way causal relationship running from broad money
lished by conducting a joint which is based on the F-test and c2 test. and domestic credit to the private sector to GDP per capita.
The long run causal relationship, on the other hand, is revealed Moreover, the results show that the total primary energy
because of the significance of the lagged error correction term in consumption played an important role in increasing both the GDP
the VECM which is based on the t test. The following equation growth and the financial development indicators in the SSA
introduces the Granger causality model with uniform lag length: countries but with a consequence of high pollution. The same
results were found by Halicioglu [2] in Turkey, Pao & Tsai [5] in the
X
l X
l BRIC countries, and Pao et al. [11] in Russia. The results also show
DDPit ¼ ait þ bit ectit1 þ xit þ DDPit þ 4it DðIDPÞit1 þ mit that the total primary energy consumption increased.
i¼1 i¼1
The conclusion arrived at by this study that the financial
(6) development cannot only affect CO2 emission and energy
DP is the dependent variable, IDP is the independent variable, D is consumption, as in Sadorsky [33], Sadorsky [34] and Zhang [32],
the first difference operator, ait is the constant term, bit, xit, and 4it, but also a long run and a positive short run bi-directional causal
are the parameters, ectit-1 is the lagged error correction term ob- relationship between total primary energy consumption, CO2
tained from the cointegrating equation and mit is the white noise emission, and the financial development indicators exist.
error. This study tends to examine whether a bi-directional causal
relationship is present between the variables. 7. Conclusion

This study investigated the impact of the energy consumption

6. Empirical results and discussion
and CO2 emission on the GDP growth and the financial develop-
ment in thirty Sub Saharan African economies, namely Benin,
Table 4 reviews the unit root test results and clearly shows that
Botswana, Burkina Faso, Burundi, Cameroon, Cape Verde, Central
the variables are not stationary at levels. However, we found that
African Republic, Chad, Comoros, Republic of Congo, Ethiopia,
the variables are stationary at the first difference rejecting the null
Gabon, Gambia, Ghana, Kenya, Lesotho, Madagascar, Malawi, Mali,
hypothesis indicating that the variables contain a panel unit root.
Mauritius, Niger, Nigeria, Rwanda, Senegal, Sierra Leone, South
Africa, Swaziland, Togo, and Zambia. The panel model was used
6.1. Panel cointegration test results taking the period 1980e2008. The results show that total primary
energy consumption and CO2 emission had a long run impact based
Table 5 reviews the Pedroni cointegration test results for the on the cointegration test results and a positive causal relationship
GDPP model, BM model and the DCPS model. From the results it based on the Granger causality test on GDP growth and the finan-
shows that most of the statistics in the GDPP model, BM model, and cial development in the investigated economies. In addition, this
the DCPS model reject the null hypothesis of no cointegration. This study found that the financial development indicators and GDP
indicates that the total primary energy consumption, CO2 emission growth had a positive causal relationship with the total primary
and investment have a long run relationship with the GDP growth. energy consumption and CO2 emission. This indicates that total
This is the same result arrived at by previous studies such as primary energy consumption increased the financial development
Menyah & Rufael [7] in South Africa and Ang [25] in Malaysia. The and GDP growth in the investigated economies with a high pollu-
results also show that the total primary energy consumption, CO2 tion consequence. The main recommendation suggested by this
emission, and investment have a long run impact on the financial study is to increase energy productivity by increasing energy
development indicators. efficiency, implementation of energy savings projects, energy
186 U. Al-mulali, C.N. Binti Che Sab / Energy 39 (2012) 180e186

conservation, and energy infrastructure outsourcing to achieve its [16] Pao H, Tsai C. CO2 emissions, energy consumption and economic growth in
BRIC countries. Energy Policy 2010;38:7850e60.
financial development and GDP growth. It is important that these
[17] Zhang X, Cheng X. Energy consumption, carbon emissions, and economic
countries increase their low investment on energy project to ach- growth in China. Ecological Economics 2009;68:2706e12.
ieve the full energy potential. [18] Menyah K, Rufael Y. Energy consumption, pollutant emissions and economic
growth in South Africa. Energy Economics 2010;32:1374e82.
[19] Pao H, Yu H, Yang Y. Modeling the CO2 emissions, energy use, and economic
References growth in Russia. Energy; 2011:1e7.
[20] Menyah K, Rufael Y. CO2 emissions, nuclear energy, renewable energy and
[1] Warr BS, Ayres RU. Evidence of causality between the quantity and quality of economic growth in the US. Energy Policy 2010;38:2911e5.
energy consumption and economic growth. Energy 2010;35:1688e93. [21] Lise W. Decomposition of CO2 emissions over 1980e2003 in Turkey. Energy
[2] Fallahi F. Causal relationship between energy consumption (EC) and GDP: Policy 2006;34:1841e52.
a markov-switching (MS) causality. Energy 2011;36:4165e70. [22] Freitas L, Kaneko S. Decomposition of CO2 emissions change from energy
[3] Payne JE. On biomass energy consumption and real output in the US. Energy consumption in Brazil: challenges and policy implications. Energy Policy
Sources, Part B: Economics, Planning, and Policy 2010;6:47e52. 2011;39:1495e504.
[4] Bownden N, Payne JE. Sectoral analysis of the causal relationship between [23] Niu S, Ding Y, Niu Y, Li Y, Luo G. Economic growth, energy conservation and
renewable and non-renewable energy consumption and real output in the US. emissions reduction: a comparative analysis based on panel data for 8 Asian-
Energy Sources, Part B: Economics, Planning, and Policy 2010;5:400e8. Pacific countries. Energy Policy 2011;39:2121e31.
[5] Yoo S. Oil consumption and economic growth: evidence from Korea. Energy [24] Lotfalipour M, Falahi M, Ashena M. Economic growth, CO2 emissions, and
Sources, Part B: Economics, Planning, and Policy 2006;1:235e43. fossil fuels consumption in Iran. Energy 2010;35:5115e20.
[6] Lee C, Chang C. The impact of energy consumption on economic growth: [25] Ang J. Economic development, pollutant emissions and energy consumption
evidence from linear and nonlinear models in Taiwan. Energy 2007;32: in Malaysia. Journal of Policy Modeling 2008;30:271e8.
2282e94. [26] Shumpeter J. The theory of economic development. Cambridge, MA, US:
[7] You J. China’s energy consumption and sustainable development: comparative Harvard University Press; 1932.
evidence from GDP and genuine savings. Renewable and Sustainable Energy [27] Goldsmith R. Financial structure and development. New York: Yale University
Reviews 2011;15:2984e9. Press; 1969.
[8] Kebede E, Kagochi J, Jolly C. Energy consumption and economic development [28] McKinnon R. Money and capital in economic development. Washington, DC:
in Sub-Sahara Africa. Energy Economics 2010;32:532e7. Brookings Institution; 1973.
[9] Yingzi N, Yuying G. Statistical and econometric analysis of the impact of [29] Shaw E. Financial deepening in economic development. New York: Oxford
China’s energy, environment on the economic development. Energy Procedia University Press; 1973.
2011;5:2358e62. [30] Jalil A, Feridun M. The impact of growth, energy and financial development on
[10] Yuan C, Liu S, Fang Z, Xie N. The relation between chinese economic devel- the environment in China: a cointegration analysis. Energy Economics 2011;
opment and energy consumption in the different periods. Energy Policy 2010; 33:284e91.
38:5189e98. [31] Tamazian A, Chousa J, Vadlamannati K. Does higher economic and financial
[11] Chang C. A multivariate causality test of carbon dioxide emissions, energy development lead to environmental degradation: evidence from BRIC coun-
consumption and economic growth in China. Applied Energy 2010;87:3533e7. tries. Energy Policy 2009;37:246e53.
[12] Halicioglu F. An econometric study of CO2 emissions, energy consumption, [32] Zhang Y. The impact of financial development on carbon emissions: an
income and foreign trade in Turkey. Energy Policy 2009;37:1156e64. empirical analysis in China. Energy Policy 2011;39:2197e203.
[13] Pao H, Tsai C. Modeling and forecasting the CO2 emissions, energy [33] Sadorsky P. Financial development and energy consumption in central and
consumption, and economic growth in Brazil. Energy 2011;36:2450e8. eastern European frontier economies. Energy Policy 2011;39:999e1006.
[14] Lean H, Smyth R. CO2 emissions, electricity consumption and output in [34] Sadorsky P. The impact of financial development on energy consumption in
ASEAN. Applied Energy 2010;87:1858e64. emerging economies. Energy Policy 2010;38:2528e35.
[15] Ozturk I, Acaravci A. CO2 emissions, energy consumption and economic growth [35] Baltagi B. Econometric analysis of panel data. England: John Wiley & Sons Ltd;
in Turkey. Renewable and Sustainable Energy Reviews 2010;14:3220e5. 2005.