Professional Documents
Culture Documents
Energy Policy
journal homepage: www.elsevier.com/locate/enpol
Viewpoint
a r t i c l e in f o
a b s t r a c t
Article history:
Received 27 March 2010
Accepted 20 July 2010
Available online 12 August 2010
The aim of this paper is to examine the existence and direction of the causal relationship between
energy consumption and output growth in the Indian cement industry for the period 197980 to
200405. The most recently developed panel unit root, a heterogeneous panel cointegration and panelbased error correction model, is applied within a multivariate framework. The empirical results conrm
a positive, long-run cointegrated relationship between output and energy consumption when
heterogeneous state effects are taken into account. We also found a long-run, bi-directional relationship
between energy consumption and output growth in the Indian cement industry for the study period,
implying that an increase in energy consumption directly affects the growth of this sector and that
growth stimulates further energy consumption. These empirical ndings imply that energy
consumption and output are jointly determined and affect each other. The empirical evidence also
suggests the implementation of energy conservation policies oriented toward improving energy-use
efciency to avoid any negative impacts of the conservation policies on the growth of this industry.
& 2010 Elsevier Ltd. All rights reserved.
Keywords:
Energy-output causality
Indian cement industry
Panel error correction model
1. Introduction
Energy is considered a basic building block for almost all
economic activities. Future economic growth crucially depends on
the long-term availability of energy in increasing quantities.
However, energy use in most cases generates undesirable
emissions with severe detrimental impacts on the environment,
including climate change. Therefore, most of the emerging
economies face dual challenges. On the one hand, for potential
future growth, countries are likely to experience a rapid growth in
energy demand; on the other hand, they have to meet their
energy requirements in an environmentally friendly way to
protect the global environment from the impacts of climate
change. These dual objectives call for the implementation of
energy conservation policies.
The implementation of energy conservation policies needs
careful investigation of the direction of causation between energy
consumption and economic growth because the direction of
causation has signicant implications on policy formulation. If, for
example, there exists unidirectional causality running from
economic growth to energy consumption, energy conservation
0301-4215/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enpol.2010.07.042
6561
1
For details of the energy efciency policies initiated by the Indian
Government, see Yang (2006).
2
This approximation of rm-level data of the industry is not perfect because
we assume that all rms in a particular state produce equally using equal amount
of inputs. In the absence of rm-level data within the states, we have used this
kind of approximation. Mukherjee (2008), in the context of Indian manufacturing,
used the same approximation.
6562
Table 1
Panel unit root testsa.
Variables
Breitung t-test
Levels
Output
Energy
Capital
Material
Labor
1.9849
0.0570
1.2034
1.8587
1.2529
First difference
Output
Energy
Capital
Material
Labor
8.7929
12.2928
9.9779
6.9659
10.3750
Probability value
(4)
(1)
(3)
(4)
(4)
0.9764
0.4772
0.3810
0.9685
0.2710
(3)n
(1)n
(1)n
(4)n
(3)n
0.0000
0.0000
0.0000
0.0000
0.0000
a
Breitung unit root test was performed with individual trends and intercepts
for each series. All variables are in natural logarithm. The number in the
parenthesis denotes optimal lag length selected automatically using the Schwarz
information criteria. The null hypothesis is a unit root.
n
denotes statistical signicance at the 1% level.
pX
1
bik Xi,tk et
k1
In Eq. (1), the Breitung (2000) test statistic examines the null
hypothesis that the process is a difference stationary:
P
H0 : pk 11 bik 1 0. The alternative hypothesis assumes that
Pp 1
the panel series is stationary, i.e.,
b 1 o 0 for all i.
k 1 ik
Breitung (2000) uses the following transformed vectors to
construct the test statistic:
=
Yi AYi yi1 ,yi2 ,. . .,yiT
=
Xi AXi xi1 ,xi2 ,. . .,xiT
These transformed vectors are then used to construct the
following test statistic:
PN
2 = =
i 1 si yi xi
2
lB q
PN
2 = =
i 1 si Xi A AXi
The statistic follows a standard normal distribution. The
results of the Breitung (2000) panel unit root test are reported
in Table 1. The test statistics for the log-levels of output, energy,
capital, labor and material are insignicant, implying that each of
3
The discussion of the Breitung panel unit root test is mostly drawn from
Narayan and Smyth (2009).
Once it is found from the unit root test that the variables are
non-stationary, i.e., they are integrated of order one, then the next
step is to apply cointegration analysis to examine whether a longrun cointegration relationship exists among those variables.
In this study, we applied Pedronis method of cointegration to
allow for heterogeneity across individual members of the panel. In
its most general form, he considered the following type of
regression:
yit ai di t bi Xit eit
Table 2
Pedroni panel cointegration tests.
Test statisticsa
No time effects
Panel
Panel
Panel
Panel
0.5037
0.1390
7.6145n
7.9684n
0.5710
1.5797
4.9733n
4.8763n
1.9308
7.8434n
6.7372n
0.9255
2.9020n
2.8462n
v
rho
PP
ADF
Group rho
Group PP
Group ADF
a
Statistics are asymptotically distributed as normal. The panel v-test is rightsided, whereas the others are left sided.
n
implies rejection of the null of no cointegration at the 1% level.
Except for the panel variance, the panel rho and the group rho
statistics, all other statistics signicantly reject the null of no
cointegration.4 Therefore, it can be inferred from the cointegration
result that a co-movement exists among output, capital, labor,
material and energy in the long run.
4.3. Panel causality test
Once the variables were cointegrated, the next step performed
was the causality test. We used a panel-based (VECM) to identify
the existence and direction of a long-run equilibrium relationship
using the two-step procedure of Engle and Granger (1987). In the
rst step, we estimated the long-run model using
ln Yit ait dit t g1i ln Eit g2i ln Kit g3i ln Lit g4i ln Mit eit
y13ik DKitk
y14ik DLitk
y21ik DEitk
y22ik DYitk
y24ik DLitk
hypothesis H0:y12ik 0 for all i and k in Eq. (5) or H0:y22ik 0 for all
i and k in Eq. (6). Masih and Masih (1996) and Asafu-Adjaye
(2000) interpreted the short run causality as a weak Granger
causality.
Another possible source of causation is the ECT in Eqs. (5) and
(6). The coefcient of the ECTs, l, is called speed of adjustment,
which represents how fast deviations from the long-run equilibrium are eliminated following changes in each variable
(Mehrara, 2007). For long-run causality, we tested H0:l1i 0 for
all i in Eq. (5) or H0:l2i 0 for all i in Eq. (6). For example, if l1i 0,
the output (Y) does not respond to a deviation from the long-run
equilibrium in the previous period. Indeed l1i 0 or l2i 0 for all i
is equivalent to both Granger non-causality in the long run and
weak exogeneity (Hatanaka, 1996 in Mehrara, 2007).
Finally, we tested whether the two sources of causation were
jointly signicant. This step was done by testing the joint
hypothesis H0:l1i 0 and y12ik 0 for all i and k in Eq. (5) or
H0:l2i 0 and y22ik 0 for all i in Eq. (6). We used a standard F-test
to test the hypothesis of joint signicance. The results of the panel
causality test are reported in Table 3.
As is clear from Table 3, the coefcients of E and ECT were
signicant at the 1% level for the output equation. Additionally,
the coefcients of Y and ECT were also signicant at the 1% level in
the energy equation. This result implied that there is both a shortrun and long-run bi-directional causality between energy consumption and output growth in the Indian cement industry.
Moreover, the ECT combined with the joint tests of E, K, L and M in
the output equation was statistically signicant at the 1% level,
implying all these variables play an important role in determining
output and correspond to theoretical expectations. Similarly, in
the energy equation, the joint tests of ECT with other variables
were signicant except for labor, implying capital and material
determine energy demand both in short run and long run.
Our nding of bi-directional causality between energy consumption and output in an Indian context is consistent with the
ndings of Paul and Bhattacharya (2004), though their study used
aggregated Indian data covering the period 19501996, whereas
ours is a disaggregated study considering only the cement
industry between 19791980 and 20042005.
y23ik DKitk
6563
6564
Table 3
Panel causality test results.
Dependent variable
Long run
DY
DE
57.49n
DY
DE
23.59
DK
n
DL
n
37.28
43.36n
11.84
6.17
ECT
DM
n
31.25
26.37n
(ECT, Y)
n
6.59
6.70n
58.37n
(ECT, E)
24.04
(ECT, K)
38.24
44.85n
(ECT, L)
(ECT, M)
nn
31.43n
28.38n
11.95
7.15
For long-run causality, we have tested the joint signicance of the error correction term ECT with lag terms of the variables. Due to space requirements, we have omitted
the D sign before the variables.
n
nn
Table 4
Fully modied OLS estimates.
Dependent variablesa
Independent variablesa
Y
Y
E
0.62 (24.39)n
a
n
E
0.34 (23.74)
K
n
L
n
0.15 (6.03)
0.13 (5.68)n
0.05 (6.48)
0.11 (1.09)
M
n
0.58 (37.64)n
0.77 (13.97)n
All variables are in natural logarithmic form. t-values are given in parentheses.
indicates signicance at the 1% level.
Acknowledgments
We gratefully acknowledge Prof. Rabindranath Bhattacharya
and Prof. M. Ramachandran for their valuable suggestions given at
different stages of preparing the paper. Thanks are also due to two
6565
Maddala, G.S., Wu, S., 1999. A comparative study of unit root tests with panel data
and a new simple test. Oxford Bulletin of Economics and Statistics, 631652.
Mandal, S.K., Madheswaran, S., 2010. Energy use efciency of Indian cement
companies: a data envelopment analysis. Energy Efciency, in press, doi:
10.1007/s12053-010-9081-7.
Masih, A.M.M., Masih, R., 1996. Electricity consumption, real income and
temporal causality: results from a multi-country study based on
cointegration and error correction modeling techniques. Energy Economics
18, 165183.
Mehrara, M., 2007. Energy consumption and economic growth: the case of oil
exporting countries. Energy Policy 35, 29392945.
Mishra, V., Smyth, R., Sharma, S., 2009. The energy-GDP nexus: evidence from a
panel of Pacic Island Countries. Resource and Energy Economics 31 (3),
210229.
Mukherjee, K., 2008. Energy use efciency in the Indian manufacturing sector: an
interstate analysis. Energy Policy 36, 662672.
Narayan, P.K., Smyth, R., 2009. Multivariate Granger causality between electricity
consumption, exports and GDP: evidence from a panel of Middle Eastern
countries. Energy Policy 37, 229236.
Paul, S., Bhattacharya, R.N., 2004. Causality between energy consumption and
economic growth in India: a note on conicting results. Energy Economics 26,
977983.
Payne, J.E., Bowden, N., 2009. The causal relationship between U.S. energy
consumption and real output: a disaggregated analysis. Journal of Policy
Modeling 31 (2), 180188.
Pedroni, P., 1999. Critical values for cointegration tests in heterogeneous panels
with multiple regressors. Oxford Bulletin of Economics and Statistics 61 (4),
549.
Pedroni, P., 2000. Fully modied OLS for heterogeneous cointegrated panels.
Advances in Econometrics 15, 93130.
Pedroni, P., 2004. Panel cointegration: asymptotic and nite sample properties of
pooled time series tests with an application to PPP hypothesis: new results.
Econometric Theory 20 (3), 597627.
Yang, M., 2006. Energy efciency policy impact in India: case study of investment
in industrial energy efciency. Energy Policy 34, 31043114.
Yoo, S-H., 2006. The causal relationship between electricity consumption and
economic growth in the ASEAN countries. Energy Policy 34, 35733582.