Page 1
European Journal of Social Sciences – Volume 16, Number 2 (2010)
206
Energy Consumption and Economic Growth: A Case Study of
Three SAARC Countries
Kashif Imran
M. Phil Scholar; Applied Economics Research Centre, University of Karachi. Pakistan
Email: k.imran_aerc@yahoo.com
Masood Mashkoor Siddiqui
Dr. Chairman, Department of Commerce
Federal Urdu University of Karachi. Pakistan
Email: drmasoodmashkoor@yahoo.com
Abstract
Energy is a crucial component to economic growth and plays a vital role in economic
development. This study inquires the causal relationships between energy consumption
(EC) and the economic growth (EG) within a multivariate framework that includes capital
stock and labor input for the panel of three SAARC countries by using modern panel unit
root technique, residual based panel cointegration and panel based error correction models
.The empirical results fully support a cointegration relationship between EC and EG in the
long run. But from the causal point of view there is long run unidirectional causality
running from EC to EG and no causality was found in the short run.
Keywords: Energy consumption, Economic growth, Causality relationship, Panel
cointegration, SAARC countries
JEL Classification Codes: O13; Q43; C33
Introduction
Literature shows that macroeconomic growth theories focus on labor and capital frequently;
researchers do not affix necessary importance to the role of energy which is important for economic
growth and production (Stern and Cleveland 2004). The economists since Adam Smith have discussed
the major inputs to economic activity as being land, labor, and capital. Neoclassical production
function explains economic growth with enlargement in labor, capital and technology; total factor
efficiency is used as technology. Growth of industrial nations in the nineteenth century can be seen in
retrospect to have been the result of fourth major input energy. Energy can also mentioned as a
production factor apart from labor and capital. By contrast, there are other energy economists who
consider that energy is a significant factor of production as well as a key player in the production
process, because it can directly be used to produce a final product (Stern 2000). All production and
many consumption activities require energy as an essential input. It is the key source of economic
growth, industrialization and urbanization.
In the light of above arguments we take the multivariate econometric model on the basis of
aggregate production function which includes labor force and capital as controlled variables. This
study inquires long run comovement and the causal relationship between energy consumption and
economic growth, for three SAARC countries like Bangladesh, India and Pakistan; collectively called
sub continent, which have a large portion of world’s population. It has more than 94% population of
South Asia, 35% of Asia and more than 21% of the world, thus the relationships between energy
Page 2
European Journal of Social Sciences – Volume 16, Number 2 (2010)
207
consumption (EC) and economic growth (EG) in such a dense populated area is the focus of this study.
We determine the relationship between EC and EG within a multivariate framework. After checking
the stationarity of variables through modern panel unit root test, we use the cointegration test which is
more powerful than cross section approach, and then use the Dynamic OLS (DOLS) technique to
estimate the cointegration vector for heterogeneous cointegrated panels. It makes possible to correct
the standard OLS for bias induced by endogeneity and serial correlation of the regressors. Furthermore,
we estimate a dynamic vector error correction model (VECM) which is suitable for heterogeneous
panels and to differentiate between short and long run causality.
We find the EC and EG relationship by employing bivariate causality procedure. But these
tests give conflicting results because of a problem often fail to detect additional channels (such as
capital and labor input). Thus, in the presence of cointegration among the series, what has been long
overdue is an alternative, superior econometric method to the vector autoregressive (VAR) method
(AsafuAdjaye, 2000). Here, we use the VECM, because the VAR models may only be able to identify
short run relationship and they are unable to determine cointegration among variables because a long
run relationship is go missing with the first differencing; while the VECM have ability of
distinguishing between short and long run relationship among variables as well as identify the sources
of causation (Oh and Lee 2004b).
Literature Review
By using of econometric techniques, the relationship between EC and EG has been analyzed for
different countries and periods by various researchers and a broad literature is available in this field.
The series between EC and EG has been the focus of extensive research for much of the past three
decades. A large number of studies show the causal relationship between both variables; conducted for
developing and developed economies. The first study in this regard was conducted by Kraft and Kraft
(1978), in their study relationship between USA’s EC and GNP for the period of 19471974 was
investigated; a unidirectional causality from GNP to EC was found. Later, Akarca and Long (1980)
tested this relationship with the same variables for same country for 1947–1972 period; unlike Kraft
and Kraft (1978) they could not found relationship between variables. Erol and Yu (1987) examined
the relationship between EC and GDP for England, France, Italy, Germany, Canada and Japan with the
data spanning 1952 to1982 and found bidirectional causality relationship for Japan, unidirectional from
EC to GDP for Canada and unidirectional from GDP to EC for Germany and Italy. They could not
found any causality relationship for France and England. A common strength of these studies is the use
of bivariate models.
Stern (1993) declared that causality relationship in bivariate models is not healthy since the
substitution effect of energy with other variables is ignored; he inquired the relationship between
USA’s EC and GDP with a multivariate cointegration model and could not found any relationship.
Ghali and ElSakka (2004) reported the short run dynamics of the variables, which indicate the
bidirectional Grangercausality between output growth and EC in the case of Canada. Various
researchers focus on panel data to investigate the causality relationship between same variables. Al
Iriani (2006) used a bivariate model for six countries making up the Gulf Cooperation Council, while
the Lee (2005) used a trivariate model with fixed capital formation for 18 developing countries.
Mehrara (2007) found unidirectional causality from EG to EC for 11 oil exporting countries. Lee and
Chang (2007) found bidirectional causality between EC and EG in case of twentytwo developed
countries while unidirectional causality from EG to EC in eighteen developing countries. Lee and
Chang (2008) found unidirectional causality running from EC to EG for Asian economies for the
period of 1971 to 2002.
Most often literature shows a relationship between EC and EG. However, there is no clear trend
in the literature. It is depending on the methodology used, the country and time period span used. It
Page 3
European Journal of Social Sciences – Volume 16, Number 2 (2010)
208
may bidirectional causality between both variables, unidirectional running from EC to EG or EG to
EC, and no causality in either direction in accordance with the ‘neutrality hypotheses’.
Data and Methodology
The econometric model used in present study is based on following production function.
Y = ƒ (EC, LF, K)
So our model is as follows;
KLFECGDP
ψψψφ+++=
321
Real GDP used for economic growth, while EC, LF, and K represent energy consumption,
labor force, and capital stock respectively. Real gross fixed capital formation used for capital stock as
proxy, as many researchers used1. All the variables are in natural log form.
Annual data used in the present study for three SAARC countries spanning the period of 1971
to 2008. The sample includes Bangladesh, India and Pakistan. Data for real GDP (2000 = 100), energy
consumption, labor force and real gross fixed capital formation (2000 = 100) are obtained from the
World Development Indicators (WDI) and annual survey reports of countries. The units are expressed
in million US$ for GDP and capital stock, numbers in million for labor force, and kilotons of oil
equivalent for energy consumption.
The first step is concerned to establish the degree of integration for each variable. So, for this
purpose this study tests the existence of unit root at level and difference for each series in the sample.
In the past decade or so, there has been much interest to test for the presence of unit root in panel data,
a number of investigators, eminent Levin, Lin and Chu [LLC (2002)], Breitung (2000), Im, Pesaran,
and Shin [IPS (2003)], and Maddala and Wu (1999)2 have involved panel based unit root tests that are
resemblance to tests carried out on a single series. These investigators have shown that panel unit root
tests are more powerful than others. Panel unit root tests lead to statistics with a normal distribution in
the limit3. This study used Im, Pesaran, and Shin [IPS (2003)], unit root test. The mention test has null
hypothesis of panel contain a unit root. Results of panel unit root test can be seen in Table: 1
Table 1: Panel Unit Root Test Results
GDP EC
Level
IPS 1.65 0.52
1st Difference
IPS 13.06* 9.10*
All variables are in natural log form *, ** and *** indicate statistical significant at 1, 5 and 10% level of significance
respectively.
Table (1) shows that variables are non stationary at level. But at first difference all the variables
are in stationary position. So, next step is to imply the cointegration test. Various researchers used
different cointegration tests for panel data e.g. King and Hillier (1985) proposed residual based LM
test. Engle and Granger (1987) also used residual based test. Pedroni (1997) and Philips and Moon
(1999) proposed a FMOLS estimator for cointegration test4. Kao (1999) propose residual based DF and
ADF for cointegration in panel data. Gutierrez (2003) and Banerjee et al. (2004) studied small sample
performance of many of these panel tests using Monte Carlo simulations, and found that no one can be
said to dominate the others. In terms of applied work, however the class of residual based tests has
ititi itiitiiit
ε+
(1)
LF
1.27
6.33*
K
0.25
8.28*
1 Paul and Bhattacharya (2004), Beaudreau (2005), Lee (2005), Thompson (2006), and Sari & Soytas (2007)
2 Maddala and Wu proposed two types of nonparametric tests including FisherADF and FisherPP statistics. In addition, for controlling
large sample size, Choi (2001) proposes two other test statistics besides Fisher’s inverse chisquare test statistics.
3 See Baltagi 2003 for detail.
4 Testing for error correction in panel data; January, 2005
Page 4
European Journal of Social Sciences – Volume 16, Number 2 (2010)
209
proven to be the most popular one5. This empirical work used residual based ADF and PP statistic for
cointegration, results are shown in Table (2).
Table 2: Residual Based Cointegration Results
Test Test statistic
ADF statistic 70.65
PP statistic 68.86
Null hypothesis: No cointegration.
Both ADF and PP test have same null hypothesis of no cointegration. Results from both tests
show that cointegration relationship exists in the variables. Thus, it can be predicted that GDP, EC, LF
and K move together in the long run. So, there is a long run steady state relationship between EC and
GDP for a cross section of selected economies.
Different researchers used different estimators to estimate panel cointegration vector. These
estimators include OLS, Fully Modified OLS (FMOLS), and Dynamic OLS (DOLS). In these
estimators DOLS has some advantages over OLS and FMOLS. It is proposed by Stock and Watson
(1993). Dynamic OLS become better than OLS by coping with small sample and energetic sources of
bias, while in FMOLS the Johansen method, being a full information technique, is exposed to the
problem that parameter estimates in one equation are affected by any misspecification in other
equations. The Stock and Watson method is by contrast a robust single equation approach which
corrects for regressors endogeneity by inclusion of leads and lags of first difference of the regressors,
and for serially correlated errors by a GLS procedure, which is not well handled by OLS. Table (3)
shows the results of the long run relationship by using DOLS estimator. All the variables in the model
are highly significant in DOLS case.
Table 3: DOLS Results
Variables
EC
LF
K
In parentheses standard error are given
* indicates statistical significant at 1%.
After concluding that all the variables in model are cointegrated, we can implement a panel
based error correction model to examine short run and long run causality between EC and economic
growth,6 because cointegration results show that causality exists between the two series but it does not
indicate the direction of causal relationship. Thus, the next step is to apply the Granger causality test.
Granger causality is a two steps procedure. The first step relates to the estimation of the
residuals
it
ε
from long run relationship. Incorporating the residuals as a right hand side variable, the
short run error correction model is calculated at the second step. Defining the error term from Equation
(1) to be
it
ECT , the dynamic error correction model of our concern is mentioned as follows:
∑∑
Ω+∆Ω++Ω=∆
−−
kk
∑∑
Ω+∆Ω+γ+Ω=∆
−−
kk
Level of Significance
0.00
0.00
DOLS
0.14* (0.042)
0.21* (0.043)
0.56* (0.049)
∑
k
∑
k
∑
k
∑
k
+∆Ω+∆Ω+∆
−−−
itkit ikkit ikk it ikkitikitiijit
K LFECGDPECTGDP
11413121111
µγ
(2)
µ+∆Ω+∆Ω+∆
−−−
it2kitik24kitik23k itik22k itik211it i 2j2it
KLFECGDPECTEC
(3)
5 See “Mixed Signals among Tests for Panel Cointegration” By Westerlund and Basher (2007).
6 The VAR models may suggest a short run relationship between variables because long run information is removed in the first
differencing, but the VECM model can evade this deficiency. Further, the VECM can recognize sources of causation and can
differentiate between a long run and a short run relationship in the series which the usual Granger causality test cannot do. Moreover,
the VAR method may not be suitable in the presence of cointegration.
Page 5
European Journal of Social Sciences – Volume 16, Number 2 (2010)
210
The sources of causation can be discovered by testing the significance of coefficients of the
dependent variables in Equations (2) and (3). First, the short run effects can be examined briefly. For
short run causality, we test 0:
120
=Ω
ik
H
for all i and k in Equation (2) or
and k in Equation (3). Next, we test long run causality by looking at the significance of the speed of
adjustment γ; the coefficient of the error correction term. The significance of γ determines the long run
relationship in the cointegrated process, and movements along this path can be considered permanent.
For long run causality, we test 0:
10
=
i
H
γ
for all i in Equation (2) or
Equation (3). Short run results of panel causality of our model are represent in Table: 4
Table 4:
Panel Causality Test Results For Short Run
0:
210
=Ω
ik
H
for all i
0:
20
=
i
H
γ
for all i in
Dependent Variables
Independent Variables
∆EC
1.29

∆GDP

1.29
∆LF
1.5
2.54
∆K
0.17
0.24
∆GDP
∆EC
Causality Results
The major goal of this study is to test the causality between EC and EG for three SAARC countries
over the period 1971 to 2008. The recently developed panel cointegration techniques are applied to
explore the relationship between the two economic series; EC and economic growth. The cointegration
tests can tell us that causality exist but can not the direction, to know the direction of causality we
employ the Granger causality test, results for short run causality are represent in Table (4), results are
showing that no variable is significant at even 10 % level of significance. So, we can conclude that
there is no short run causality exists in either direction; from GDP to EC or from EC to GDP. To know
the results for long run causality we see the coefficient of Error Correction Term (ECT), if the
coefficient is significant and has negative sign then long run relationship exist between variables
otherwise not, and ECT coefficient value is (0.041) in our results and significant at 5% level of
significance. So we can conclude that in the long run there is causality exist among variables running
from EC to economic growth. But the coefficient is insignificant in reverse case; when EC take as
dependent variable. So in this case no causality exists. It means a unidirectional causality exists in our
model. In other words we can say that energy treat as an engine of accelerated economic growth and
the changes in EC have a significant effect on growth. Our results are fully consistent with those of
Stern (2000), Oh and Lee (2004a) and Paul and Bhattacharya (2004) who point out energy as one of
the cornerstones of the aggregate production function in terms of supply; however compared with
previous empirical studies, we note that the findings also support those of Yu and Choi (1985) for the
Philippines, Masih and Masih (1998) for Thailand as well as those of AsafuAdjaye (2000), Fatai et al.
(2004) for Indonesia and Lee and Chang (2008) for Asian economies. Thus, there is every reason to
believe that the results also deny the neutrality hypothesis for the energyGDP relationship in three
SAARC economies. For mentioned countries, in general it is clear that energy is an important
component for economic growth in the long run. In fact, production in industries and agriculture also,
really demands a sturdy amount of energy. Important too, the energy equations are not significant
when EC is dependent variable, which indicates that there is no long run causal relationship running
from GDP to EC.
Conclusions and Policy Implications
This study aims to determine the causal relationship between energy consumption (EC) and economic
growth (EG) in the case of three SAARC countries i.e. Bangladesh, India, and Pakistan. Since the
energy is an important factor in production so in this paper we used the production side model to
Page 6
European Journal of Social Sciences – Volume 16, Number 2 (2010)
211
empirically reinvestigate the causal relationship between EC and EG in a multivariate model by using
the data from 1971 to 2008. Time series data may give inconsistent and unpredictable results with the
short time period of typical data sets, This study applies new heterogeneous panel cointegration and
panel based error correction model techniques to reexamine the relationship between EC and real
GDP among mentioned countries.
On the basis of our short run and long run results we can reject the neutrality hypothesis that
has previously been advanced. EC is found to Granger cause GDP in the long run, but not in short run.
No any causal relationship exists running from GDP to EC or EC to GDP in short run, but exists in
long run running from EC to EG. So, we can say high EC tends to come with high GDP, but not the
reverse. In the light of above discussion it is reflecting that energy serves as an engine of economic
growth and economic activity will be affected in the result of changes in EC. This means that
continuous energy use does produce a continuous increase in output. Thus, GDP is basically
determined by energy, but while energy conservation may be feasible. So the related authorities in
SAARC economies should take a special interest in different sources of energy and invest more in this
sector, and invite foreign investors to invest in this sector, make suitable policies in this regard and find
new alternate and cheap sources of energy like other develop economies. More water reservoirs are
needed in three mentioned economies to produce energy at cheap cost and at large scale. Improvement
in or establishment of R&D departments and increase their efficiency is also need of time, so that it
create multiplier effect on GDP and as a result prosperity will come in these economies. Short fall of
energy sources may harm the economic growth and push back the economy in several ways. Through
effective changes and better policies in this channel can change the living standards of people in
developing economies like Bangladesh, India, and Pakistan.
References
[1]
Akarca,A.T and Long,T.V (1980) “On the Relationship between Energy and GNP: A Re
examination”, Journal of Energy and Development, Vol. 5, pp. 32631.
[2]
AlIriani,M.A (2006) “Energy–GDP relationship revisited: an example from GCC countries
using panel causality”, Energy Policy Vol. 34 (17) pp. 3342–50.
[3]
AsafuAdjaye,J (2000) “The relationship between energy consumption, energy prices and
economic growth: time series evidence from Asian developing countries”, Energy Economics
Vol. 22, pp. 615–625.
[4]
Baltagi, H.B. (2003) “Econometric Analysis of Panel Data”, 2nd Edition, John Wiley & Sons
Ltd, New York.
[5]
Banerjee,Anindya, Marcellino,Massimiliano and Osbat,Chiara (2004) “Some Cautions on the
Use of Panel Methods for Integrated Series of Macroeconomic Data”, Econometrics Journal,
Vol. 7(2) pp. 32240.
[6]
Beaudreau,  B.C (2005) “Engineering and economic growth”, Structural Change and
Economic Dynamics, Vol. 16, pp. 211– 20.
[7]
Breitung,J (2000) “The local power of some unit root tests for panel data. In: Baltagi, B.,
Fomby, T.B., Hill, R.C. (Eds.), Advances in Econometrics: Nonstationary Panels”, Panels
Cointegration and Dynamic Panels Vol. 15, pp. 161–78.
[8]
Choi,I. (2001) “Unit root tests for panel data”, Journal of International Money and Finance,
Vol. 20, pp. 24972.
[9]
Engle,R. and Granger,C (1987) “Cointegration and error correction: representation,
estimation, and testing”, Econometrica, Vol. 55, pp. 257–76.
[10]
Erol,U and Yu,E.S.H (1987) “Time series analysis of the causal relationships between U.S.
energy and employment”, Resources and Energy, Vol. 9, pp. 75–89.
Page 7
European Journal of Social Sciences – Volume 16, Number 2 (2010)
212
[11]
Fatai,K, Oxley,L and Scrimgeour,F.G (2004) “Modelling the causal relationship between
energy consumption and GDP in New Zealand, Australia, India, Indonesia, the Phillippines and
Thailand”, Mathematics and Computers in Simulation, Vol. 64, pp. 431– 45.
Ghali,K.H, and ElSakka,M.I.T (2004) “Energy use and output growth in Canada: a
multivariate cointegration analysis” Energy Economics, Vol. 26 (2), pp.225–38.
Gutierrez,Luciano (2003) “On the Power of Panel Cointegration Tests: A Monte Carlo
Comparison”, Economics Letters, Vol. 80(1), pp. 10511.
Harris,R.D and Tzavalis,E (1999) “Inference for unit roots in dynamic panels where the time
dimension is fixed”, Journal of Econometrics, Vol. 91 (2), pp. 201–26.
Im,K.S, Pesaran,M.H and Shin,Y (2003) “Testing for unit roots in heterogeneous panels”,
Journal of Econometrics Vol.115, pp. 53– 74.
Kao,Chihwa (1999). “Spurious Regression and ResidualBased Tests for Cointegration in
Panel Data”, Journal of Econometrics, Vol. 90(1), pp. 144.
King,M.A and Hiller,G.H. (1985), “Locally Best Invariant Tests of the Error Covariance
Matrix of the Linear Regression Model”, Journal of the Royal Statistical Society, Vol. 47, pp.
98102.
Kraft,J and Kraft,A (1978) “On the relationship between energy and GNP” Journal of Energy
and Development, Vol. 3, pp. 401–03.
Lee,C.C (2005) “Energy consumption and GDP in developing countries: a cointegrated panel
analysis” Energy Economics, Vol. 27, pp. 415–27.
Lee,C.C and Chang,C.P (2007) “Energy consumption and GDP revisited: A panel analysis of
developed and developing countries”, Energy Economics, Vol.29 (2007) pp.1206–23.
Lee,C.C and Chang,C.P (2008) “Energy consumption and economic growth in Asian
economies: A more comprehensive analysis using panel data”, Resource and Energy
Economics, Vol. 30, pp. 50–65
Levin,A, Lin,C.F and Chu,C.S (2002) “Unit root tests in panel data: asymptotic and finite
sample properties”, Journal of Econometrics, Vol. 108 (1) pp. 1–24.
Maddala,G.S and Wu,S.A (1999) “Comparative study of unit root tests with panel data and a
new simple test”, Oxford Bulletin of Economics and Statistics, Vol. 61, pp. 631–52.
Mehrara,Mohsen (2007) “Energy consumption and economic growth: The case of oil
exporting countries”, Energy Policy, Vol. 35, pp. 2939– 45.
Masih,A.M.M and Masih,R (1998) “A multivariate cointegrated modeling approach in testing
temporal causality between energy consumption, real income and prices with an application to
two Asian LDCs” Applied Economics, Vol.30 (10), pp. 1287–98.
Oh,W and Lee,K (2004a) “Causal relationship between energy consumption and GDP
revisited: the case of Korea 1970– 1999”, Energy Economics Vol. 26, 51–59.
Oh,W and Lee,K (2004b) “Energy consumption and economic growth in Korea: testing the
causality relation” Journal of Policy Modeling, Vol. 26, pp. 973–81.
Paul,S and Bhattacharya,R.N (2004) “Causality between energy consumption and economic
growth in India: a note on conflicting results”, Energy Economics, Vol. 26, pp. 977–83.
Pedroni,P (1997) “Panel Cointegration; Asymptotic and Finite Sample Properties of Pooled
Time Series Tests, with an Application to the PPP Hypothesis: New Results”, working paper,
Indiana University.
Peter,CBPhillips and Hyungsik,RMoon (1999). “Linear Regression Limit Theory for
Nonstationary Panel Data”, Econometrica, Vol. 67(5), pp. 10571112.
Sari,R and Soytas,U. (2007) “The growth of income and energy consumption in six
developing countries”, Energy Policy Vol. 35 (2) pp. 889–98.
Stern,D. I (1993) “Energy and Economic Growth in the USA, A Multivariate Approach”,
Energy Economics Vol.15, pp. 13750.
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
Page 8
European Journal of Social Sciences – Volume 16, Number 2 (2010)
213
[33]
Stern,D. I (2000) “Multivariate cointegration analysis of the role of energy in the U.S.
macroeconomy”, Energy Economics, Vol. 22, pp. 267–83.
Stern,D. I and Cleveland,C.J (2004) “Energy and Economic Growth”, Rensselaer Working
Papers in Economics 0410, Rensselaer Polytechnic Institute, Department of Economics.
Stock,J. H. and Watson,M. W. (1993) “A simple estimator of cointegrating vectors in higher
order integrated systems”, Econometrica, Vol. 61, pp. 783–820.
Thompson,H. (2006) “The applied theory of energy substitution in production”, Energy
Economics Vol. 28 (4), pp. 410–25.
Yu,S.H and Choi,J.Y (1985) “The causal relationship between energy and GNP: an
international comparison”, Journal of Energy Development, Vol. 10, pp. 249–72.
Westerlund,J. and Basher,S. A. (2008). “Mixed Signals Among Tests for Panel
Cointegration”, Economic Modeling, Vol. 25(1), pp. 12836
[34]
[35]
[36]
[37]
[38]