67736459

European Journal of Social Sciences ISSN 1450-2267 Vol.25 No.3 (2011), pp. 316-328 © EuroJournals Publishing, Inc. 2011 http://www.europeanjournalofsocialsciences.com

316

Does Electricity Consumption Panel Granger Cause Economic

Growth in South Asia? Evidence from

Bangladesh, India, Iran, Nepal, Pakistan and Sri-Lanka

Md. Sharif Hossain

Associate Professor of Econometrics

Department of Economic Engineering

Faculty of Economics, Kyushu University, Japan

E-mail: [email protected], [email protected]

Chikayoshi Saeki

Professor of Econometrics

Department of Economic Engineering

Faculty of Economics, Kyushu University, Japan

E-mail: [email protected]

Abstract

This paper empirically examines the dynamic causal relationships between electricity consumption and economic growth for the panel of south Asian countries using time series data from 1971 to 2007. Another purpose of this paper is to find the short-run and long-run elasticities of economic growth with respect to electricity consumption for this panel. The four panel unit root tests results support that both the variables are integrated of order 1. The Johansen Fisher panel conintegration test results support that there is a one cointegration vector. The Granger causality tests results support that existence of unidirectional causality from economic growth to electricity consumption in India, Nepal and Pakistan, and from electricity consumption to economic growth in Bangladesh. No causal relationship is found in Iran and Sri-Lanka. The panel Granger F-test results support that there is no evidence of short-run causal relationship between the variables but long-run unidirectional causal relationship from electricity consumption to economic growth is found for this panel. It is found that the long-run elasticity of economic growth with respect to electricity consumption (0.5451) is higher than short-run elasticity of 0.3813. This means that over times higher electricity consumption in South Asian countries gives rise to more economic growth. Keywords: Dynamic causal relationship. Panel unit root test, Panel cointegration test,

Granger causality test. JEL Classification Codes: C23, C32, C33, O50, O57, Q40

1. Introduction Due to rising energy demand of some rising Asian countries like as India, China, Philippines, Indonesia, Malaysia, soaring oil prices, concerns about energy supply security, the debate of rising GHGs and climate change, a common Asian energy policy will become indispensable for future or

European Journal of Social Sciences – Volume 25, Number 3(2011)

317

near future. Now-a-days, energy efficiency measures will play a vital role as energy savings as a result most of the developing countries fear that such policy measure will harm their economic development. Due to some Asian rising countries, Asia will be one of the most active regions in terms of renewable energy and climate policy targets in future. Thus the most import question arises whether the new energy policy and policy for reducing the GHG’s emissions will strike the economy of South Asian countries. One of the best known methods is to investigate the short-run and long-run causal relationships between energy consumption and economic growth for the panel of South Asian countries using the time series data. For this study, the variable electricity consumption (kWh per capita) and per capita real GDP (constant 2000 US $) have been considered as the proxies for energy consumption and economic growth respectively for this panel. On the basis of the modern econometrics techniques, the dynamic causal relationship between electricity consumption and economic growth has been examined in this paper. The testing procedure involves the following steps. At the first step whether each variable contains a unit root will be examined. If the variables contain a unit root, the second step is to test whether there is a long-run cointegration relationship between the variables. If a long-run relationship between the variables is found, the final step is to estimate panel vector error correction model in order to infer the Granger causal relationship between the variables. Finally using an appropriate method the long-run and short-run elasticities of economic growth with respect to electricity consumption will be estimated. The direction and policy implications for the causal relationship between electricity consumption and economic growth can be classified as follows. If the unidirectional causal relationship from electricity consumption to economic growth is found, then any restriction on the use of energy leads to a reduction of economic growth. Thus about this negative effect on economic growth that caused by a policy of restriction of energy use in order to slow down the rate of climate change grows by reducing GHG’s, many Asian countries will be worried. On the other hand if unidirectional causal relationship from economic growth to electricity consumption is found, then any restriction on the use of electricity has very little or no adverse impacts on economic growth. A bi-directional causal relationship implies that electricity consumption and economic growth are jointly determined and will affect at the same time. If no causal relationship between these two variables is found, then the hypothesis of neutrality holds indicates that any restriction on energy use will not work as a barrier for economic development of this panel. That is why in this paper the principal purpose has been made to investigate the dynamic causal relationships between economic growth and electricity consumption for the panel of south Asian countries to know the impact of new energy policy and policy for reducing the GHG’s emissions on economic growth in south Asian society. The organizational structure of the paper is as follows: Section 2 discusses the literature review; Section 3 discusses data sources and descriptive statistics; Section 4 provides econometric modeling framework with empirical analysis and Section 5 concludes with a summary of the main findings and policy implications.

2. Literature Review In the last three decades, the causal relationship between energy consumption consumption and economic growth as well as economic growth and carbon dioxide emissions has been investigated widely in economic literature. The enormous amount of empirical literatures to examine the causal relationship between energy consumption and economic growth particularly in developed and developing countries fall into four categories (i) no causal relationship between energy consumption and economic growth, (ii) unidirectional causality from energy consumption to economic growth, (iii) unidirectional causality from economic growth to energy consumption and (iii) bidirectional causality between energy consumption and economic growth. For instance, studies that found no causal relationship between energy consumption and economic growth are as; Akarca and Long (1980), Yu and Hwang (1984), Yu and Choi (1985), Erol and Yu (1988), Stern (1993), and Cheng (1999), and Imran (2010). Studies that found unidirectional causality from economic growth to energy

European Journal of Social Sciences – Volume 25, Number 3 (2011)

318

consumption are as; Kraft and Kraft (1978), Cheng and Lai (1997), Glasure and Lee (1998), Cheng (1999), Chang and Wong (2001), Soytas and Sari (2003)) and Narayan and Smyth (2009),

A number of studies have found the direction of causality from energy consumption to economic growth are as; Yu and Choi (1985), Masih and Masih (1996), Asafu and Adjaye (2000), Yang (2000), Soytas and Sari (2003), Morimoto and Hope (2004), Shina and Lam (2004), Narayan and Singh (2007)), and Squalli (2007). The Studies that found two way causation are as; Masih and Masih (1997), Asafu and Adjaye (2000), Glasure (2002)), and Oh and Lee (2004). In time series econometrics most recent studies have tended to focus on VAR and VEC models and cointegration approach. For example Masih and Masih (1996) used the cointegration analysis to study the causal relationship between energy consumption in a panel of six Asian countries and found cointegrated relationship between these variables in India, Pakistan and Indonesia but no integration is found in Malaysia, Singapore and Philippines. The direction of causality is found from energy consumption to GDP in India, and from GDP to energy consumption in Pakistan and Philippines. Asafu and Adjaye (2000) investigated the causal relationship between energy use and income in four Asian countries using cointegration and error correction mechanism. He found that causality runs from energy use to income in India and Indonesia and bi-directional causality in Thailand and Philippines. Yang (2000), found bi-directional causality between energy consumption and GDP in Taiwan and this results contradicts with Cheng and Lai (1997) results. Soytas and Sari (2003) found bidirectional causality in Argentina and unidirectional causality from GDP to energy consumption in Italy and South Korea, and from energy consumption to GDP in Turkey, France, Germany and Japan. Paul and Bhattacharya (2004) found bidirectional causality between energy consumption and economic growth in India. Using cointegration analysis Wietze and and Van (2007) found that unidirectional causality from GDP to energy consumption in Turkey. Dirck (2008) used the cointegartion approach to study the causal relationship between electricity consumption and economic growth for the panel of 15 European countries. He found the cointegration in Great Britain, Greece, Ireland, Italy, and Netherlands, and no cointegration is found in Austria, Belgium, Germany, Denmark, Spain, Finland, France, Luxembourg, Portugal, and Switzerland. He also found the unidirectional causality from electricity consumption to economic growth for Greece, Italy, and Belgium, and from economic growth to electricity consumption for Great Britain, Ireland, Netherland, Spain and Portugal, no causality is found in Austria, Germany, Denmark, Finland, France, Luxembourg, and Switzerland. Narayan, Narayan and Popp (2010) used the cointegartion approach to study the causal relationship between electricity consumption and economic growth for six different panels of 93 countries. They found bidirectional causality relationship between these two variables except for the panel of Middle East. Unidirectional causality from GDP to electricity consumption is found for the panel of Middle East. Thus the existing literature reveals that due to the application of different econometric methodologies and different sample sizes, the empirical results are very mixed and even vary for the same country and are not conclusive to present policy formulation that can be applied over the countries especially for south Asian countries. According to the knowledge still now no one has conducted any research to find the causal relationship between electricity consumption and economic growth for the panel of South Asian countries. Thus this study for the panel of South Asian economy tries to overcoming the shortcoming literature related with the linkage between electricity consumption and economic growth. Also this empirical study will be important to formulate policy recommendation from the point of view of electricity consumption and economic growth for South Asian countries.

3. Data Sources and Descriptive Statistics Annual data for electricity consumption (EC) (kWh per capita), and per capita GDP (PGDP) (constant 2000 US $), are downloaded from the World Bank’s Development Indicators. The data is for the period from 1971 to 2007. The south Asian countries Bangladesh, India, Iran, Nepal, Pakistan and Sri-


319

Lanka have been considered for this panel analysis. Now for comparison about the variability of these variables among these countries at first the descriptive statistics are recorded below in Table (1) Table 1: Descriptive Statistics for the Individual and also for Panel

Country Electricity Consumption (EC) Per Capita Per Capita Real GDP (PGDP)

Mean Std. Dev. CV Mean Std. Dev. CV

Bangladesh 55.863292 42.131995 75.4198 277.26249 63.963244 23.0696 India 272.7620 134.4489 49.2917 341.5815 130.6930 38.2612 Iran 1053.4002 586.3068 55.6585 2057.2445 906.15418 44.0470 Nepal 35.030552 23.867642 68.1338 179.71102 67.486351 37.5527 Pakistan 259.16993 127.18071 49.0723 437.690065 112.837963 25.7803 Sri-Lanka 181.18399 105.90196 58.4500 621.095147 228.222301 36.7451 Panel 309.5683 428.045 138.2717 652.4308 750.96624 115.1028

The mean electricity consumption recorded is highest for Iran followed by India, Pakistan, Sri-

Lanka, Bangladesh and Nepal. It has also been found that Bangladesh is most volatile country in respect of electricity consumption followed by Nepal, Sri-Lanka, Iran, India and Pakistan. Based on per capita GDP, the highest per capita GDP is for Iran followed by Sri-Lanka, Pakistan, India Bangladesh and Nepal. The most volatile country in respect of per capita GDP is Iran, followed by India, Nepal, Sri-Lanka, Pakistan and Bangladesh. It has been found that the consumption of electricity of low income countries is smaller with compared to high income countries. Thus a question arises in our mind, is there any causal relationship between electricity consumption and economic growth in South Asian countries. For the panel data it has been found that average per capita electricity consumption is 309. 5683 kWh and the average per capita real income is 652.4308 USD.

4. Econometric Methodology The empirical investigation of the dynamic causal relationship between electricity consumption and economic growth involves the following three steps. At the first step whether each variable contains a unit root has been examined by using ADF test as well as the different panel unit root tests. If the variables contain a unit root, the second step is to test whether there is a long run-cointegration relationship between the variables. If a long-run relationship between the variables is found, the final step is to estimate panel vector error correction model in order to infer the Granger causal relationship between the variables if not then VAR model will be estimated. Finally using an appropriate method the long-run and short-run elasticities of economic growth with respect to electricity consumption are obtained. 4.1. Panel Unit Root Tests

We know most of the macroeconomic variables tend to exhibit a trend over time are non stationary would lead to the problem spurious regression. If economic time series are non-stationary process, the Augmented Dickey Fuller (ADF) test in order to test the unit root of order one for individual country takes the following regression equation

p

t 0 1 t-1 i t-i t

i=1

y = + t + + y +yα α θ φ ε∆ ∆∑ (1)

Here yt is the series under investigation of country i, ∆ stands for first difference and the

lagged difference terms on the right hand side of the equations are designed to correct for serial correlations of the disturbance terms. The lagged differences are selected by using the AIC and SBIC

criteria. If θ = 0, the series ty contains a unit root and therefore an I(1) process governed by a

stochastic trend. We know if the sample size is small, the traditional unit root tests are known have


320

limited power as against alternative hypothesis. Thus the panel unit root tests can eliminate the problems associated with the lower power of the traditional unit root tests, as pooling increases the sample size considerably allowing for higher degrees of freedom and hence more accurate and reliable statistical tests. Since none of the panel unit root test is free from some statistical shortcomings, in terms of size and power properties, so it is better for us to perform several unit root tests to infer an overwhelming evidence to determine the order of integration of the variables. In this paper four panel unit root tests: Levin, Lin and Chu (LLC, 2002) , Im, Peasaran and Shin (IPS, 2003), Maddala and Wu ( MW, 1999), and Choi (2006) tests have been applied. The LLC test is based on the assumption that

the persistence parameters iρ are common across cross-sections so that i = ρ ρ for all i. But this

assumption is not true for several variables. The second and third tests assume cross-sectional independence. This assumption is likely to be violated for the income variable. It has been found by Banerjee, Cockerill and Russell (2001) that these tests have poor size properties and have a tendency to over-reject the null hypothesis of unit root if the assumption of cross-section independence is not satisfied. Pesaran (2007) and Choi (2006) derived other tests statistics to solve this problem.

Levin, Lin and Chu (LLC, 2002) considered the following regression equation ip

it it-1 ij it-j it it

j=1

y = y + y +X +α γ δ ε′∆ ∑ (2)

where it it i,t-1y = y -y∆ , here the assumption is = -1α ρ i.e. i = ρ ρ for all i, but allow the lag order for the

difference terms ip , to vary across cross-sections. Here the null hypothesis to be tested is 0H : 0α = ;

against the alternative hypothesis is that 1H : < 0 α .The null hypothesis indicates that there is a unit

root while the alternative hypothesis indicates that there is no unit root. To perform the test statistics at

first they regress ity∆ and it-1y on the lag terms it-jy∆ (j = 1, 2,….., ip ) and the exogenous variables

itX which are given by;

ip

it ij it-j it it

j=1

y = y +X +uγ δ′∆ ∆∑ (3)

ip

it-1 it-j it it

j=1

y = y +X +vij

β λ′∆∑ (4)

The estimated equations are given by; ip

it ij it-j it

j=1

ˆˆy = y +Xγ δ′∆ ∆∑ (5)

ip

it-1 it-j it

j=1

ˆ ˆy = y +Xij

β λ′∆∑ (6)

Then they define ity∆ by taking ity∆ and removing the autocorrelations and deterministic

components using the first set of auxiliary estimates: ip

it it ij it-j it

j=1

ˆˆy = y y -Xγ δ′∆ ∆ − ∆∑ (7)

Analogously we may define ip

it-1 it-1 it-j it

j=1

ˆ ˆy = y y -Xij

β λ′− ∆∑ (8)

The proxies are obtained by standardizing both ity∆ and it-1y dividing by the regression

standard error: itit

i

yy

s

∆∆ =� ; and it-1

it-1

i

yy =

s� ; where is are estimated the standard errors from estimating


321

each ADF in equation (2) . Finally an estimate of the coefficient α may be obtained from the pooled proxy equation

it it-1 ity yα η∆ = +� � (9)

LLC show that under the null hypothesis, a modified t-statistic for the resulting α is asymptotically normally distributed

-2 *

n* mT

*

mT

ˆˆt (nT)S se( ) t = ~N(0, 1)α

α

σ α µ

σ

− �

�

� (10)

Where tα is the standard t-statistic for 0H : = 0 α , 2σ is the estimate variance of the error term

η , ˆse( )α is the standard error of α , nS is the mean of the ratios of the long-run standard deviation to

the innovation standard deviation for each individual. Its estimate is derived using kernel-based techniques

i

1

p

T = T- -1n

i=

∑�

(11)

*

mTµ� and

*

mTσ

� are the two adjusted factors for the mean and standard deviation.

Im, Pesaran and Shin (2003) proposed the test statistics using the following model; ip

it it-1 ij it-j it it

j=1

y = y + y +X +i

α γ δ ε′∆ ∑ (12)

where, it it i,t-1y = y -y∆ , ity (i = 1, 2,………..,n; t = 1, 2,………..,T) is the series under investigation for

country i over period t, ip is the number of lags in the ADF regression and the itε errors are assumed

to be independently and normally distributed random variables for all i’s and t’s with zero mean and

finite heterogeneous variance 2

i σ . Both iα and ip in equation (12) and are allowed to vary across

countries. The null hypothesis to be tested is that each series in the panel contains a unit root, i. e.

0 iH : = 0 iα ∀ . Against the alternative hypothesis that some of the individual series to have unit root

but not all

i

1

i

0; for some i's H :

0;for at least one i

α

α

=

< There are two stages for constructing the t-bar statistic which is proposed by Im, Pesaran and

Shin (2003). At the first stage the average value of the individual ADF t-statistic for each of the countries in the sample is calculated which is given by

i

n

nT iT i

i=1

1t = t (p )

n∑ (13)

Where iiT it (p ) is the calculated ADF test statistic for country i of the panel (i = 1, 2, ……,n).

The second step is to calculate the standardized t-bar statistic which is given by;

nT

1nT iT i

1

t

iT i

1

n t E( t (p ))

Z = ~ N(0, 1)1

var( t (p ))

n

n

i

n

in

=

=

−

∑

∑ (14)

Where n is the size of the panel, which indicates the no. of countries, iT iE(t (p )) and iT ivar( t (p ))

are provided by IPS for various values of T and p. However, Im, et al. (2003) suggested that in the presence of cross-sectional dependence, the data can be adjusted by demeaning and that the standardized demeaned t-bar statistic converges to the standard normal in the limit.


322

Maddala and Wu (1999) proposed a Fisher-type test which combines the p-values from unit root tests for each cross-section i. The test is non-parametric and has a chi-square distribution with 2n degrees of freedom, where n is the number of countries in the panel. The test statistic is given by;

n2

e i 2n(d.f.)

i=1

=-2 log (p )~λ χ∑ (15)

where ipis the p-value from the ADF unit root tests for unit i. The Maddala and Wu (1999) test has the

advantage over the Im, et al. (2003) test that it does not depend on different lag lengths in the individual ADF regressions. Maddala and Wu (1999) performed Monte Carlo simulations showing that their test is superior to that proposed by Im, et al. (2003). In addition Choi (2006) derived another test statistic which is given by;

n-1

i

i=1

1Z = (p ) ~ N(0, 1)

nΦ∑ (16)

where, -1Φ is the inverse of the standard normal cumulative distribution function. For panel unit root tests, two cases have been considered. In case one both constant and trend

terms are included in the equation and in case two only constant term is included in the equation. We know macroeconomic variables tend to exhibit a trend over time. As a result it is more appropriate to consider the regression equation with constant and trend terms at level form. Since first differencing is likely to remove any deterministic trends in the variables, regression should include only constant term. But for comparison, both constant and trend terms are considered for the test statistics while utilizing level form and first differenced of the variables in their logarithmic form. The test results for individual country and also for panel are given below in Table (2) and (3) Table 2: ADF Unit Root Test Results for the Individuals

Country Case 1 [Level form] Case 2 [Level form] Case 1[Differenced form] Case 2 [Differenced form

LnEC LnPGDP LnEC LnPGDP LnEC∆ LnPGDP∆ LnEC∆ LnPGDP∆ Bangladesh

-2.32486 1.17428 -0.32063 2.57729 -7.1952** -11.433** -7.35694** -5.46914**

India -1.49443 0.04704 -0.88156 3.49918 -3.76229* -4.61474** -3.24883* -2.96550* Iran -4.19822* -2.60034 -0.23146 -2.68655 -4.65142** -2.57894 -4.96788** -3.26096* Nepal -0.69430 -2.03944 -2.81768 -0.17622 -5.58849** -5.43540** -4.54707** -3.32439* Pakistan -0.71656 -1.87082 -1.86254 -0.81803 -3.77206* -2.56388 -3.52459* -2.59674 Sri-Lanka -2.89897 -1.65375 0.48568 1.29787 -5.01998** -3.05916 -5.00021** -2.76113

Table 3: LLC, IPS, MW and Choi Panel Unit Root Tests Results

Case 1 [Level Form]

LLC Test Prob. IPS Test Prob. MW Test Prob. Cho Test Prob.

LnPGDP 2.7568 0.9971 2.8841 0.9980 4.7970 0.9644 2.87865 0.9980 LnEC -0.7392 0.2299 -1.3595 0.0870 4.7794 0.9564 -0.94826 0.1715

Case 2 [Level Form]

LnPGDP 4.0968 1.0000 6.0818 1.0000 5.6297 0.9336 5.3373 1.0000 LnEC -2.0925* 0.0182 1.4143 0.9214 5.2941 0.9474 1.4831 0.9310

Case 1 [Differenced Form]

DLnPGDP

-3.801** 0.0001 -8.276** 0.0000 138.193** 0.0000 -7.5468** 0.0000

DLnEC -5.079** 0.0000 -7.8462** 0.0000 76.3670** 0.0000 -6.8320** 0.0000

Case 2 [Differenced Form]

DLnPGDP

-2.6839** 0.0036 -6.0422** 0.0000 61.4107** 0.0000 -5.6916** 0.0000

DLnEC -6.0145** 0.0000 -8.8063** 0.0000 91.9893** 0.0000 -7.8001** 0.0000

**: Indicates significant at 1% level, *: indicates significant at 5% level.


323

The individual unit root tests support that for the south Asian countries the null hypothesis of unit root cannot be rejected for both variables in their level form but the null hypothesis will be rejected in their differenced form. Thus it can be said that both variables are integrated of order 1. All the panel unit root tests results support that both the variables are integrated of order 1. 4.2. Panel Cointegration

From the panel unit root test it has been found that all of the series are integrated of order (1), therefore we applied the cointegration analysis to examine whether there is a long-run relationship between the variables using the Johansen Fisher panel cointegration test proposed by Maddala and Wu (1999). The Johansen Fisher panel conintegration tests is panel version of the individual Johansen conintegration test. The Johansen Fisher panel cointegration test is based on the aggregates of the p-values of the

individual Johansen maximum eigenvalues and trace statistic. If ip is the p-value from an individual

cointegration test for cross-section i, under the null hypothesis for the panel n

2

i 2

1

-2 log(p ) ~n

i

χ=

∑ (17)

The 2χ value is based on p-values for Johansen’s cointegration trace test and maximum

eigenvalue test. In the Johansen type panel cointegration tests results heavily depends on the number of lags of the VAR system. The results are obtained here for the appropriate lags of VAR system are given below in Table (4) Table 4: Results of the Cointegration Tests for Individuals

Country

Model 1: Hypothesis of No Cointegration Model 2: Hypothesis of No Cointegration

Trace

Test Prob

Max-Eigen

Value Test Prob

Trace

Test Prob

Max-Eigen

Value Test Prob

Bangladesh

20.3276 0.0086 20.3211 0.0049 33.4155 0.0047 23.6935 0.0111

India 22.7562 0.0034 22.7562 0.0018 29.7186 0.0158 23.0420 0.0140

Iran 9.6628 0.3076 8.9810 0.2877 33.4767 0.0047 30.0768 0.0010

Nepal 16.6516 0.0315 16.0465 0.0259 20.0176 0.2250 16.0843 0.1416

Pakistan 10.6516 0.2338 10.6399 0.1732 14.6116 0.6069 11.5080 0.4623

Sri-Lanka 11.6966 0.1721 9.7950 0.2257 16.3846 0.4621 14.1411 0.2448

Model 1: Hypothesis of at most One Model 2: Hypothesis of at Most One

Cointegration Equation Cointegration Equation

Bangladesh

0.0065 0.9352 0.0065 0.9352 9.7220 0.1404 9.7220 0.1404

India 0.0051 0.9969 0.0051 0.9969 6.6765 0.3795 6.6765 0.3795

Iran 0.6818 0.4090 0.6818 0.4090 3.3999 0.8260 3.3999 0.8260

Nepal 0.7691 0.3805 0.7691 0.3805 3.9334 0.7517 3.9334 0.7517

Pakistan 0.0118 0.9134 0.0118 0.9134 3.1036 0.8640 3.1036 0.8640

Sri-Lanka 1.9016 0.1679 1.9016 0.1679 2.2435 0.9518 2.2435 0.9518

Table 5: Results of the Johansen Based Panel Cointegration Test

Number of

Coint. Eqn.

Model 1 Model 2

Trace

Test Prob.

Max-Eigen

Value Test Prob.

Trace

Test Prob.

Max-Eigen

Value Test Prob.

None 36.59 0.0003 39.54 0.0001 35.27 0.0004 39.68 0.0001 At Most 1 7.611 0.8148 7.611 0.8148 7.209 0.8435 7.209 0.8435

From the estimated results in table (4) it has been found that there is a long-run relationship

between electricity consumption and per capita GDP for four countries Bangladesh, India, Iran and


324

Nepal out of 6 countries. Thus to investigate the causal relationship between these variables, VEC (vector error correction) models have been considered for Bangladesh, India, Iran and Nepal but for Pakistan and Sri-Lanka VAR models have been used to find the causal relationship between these two variables. From the Johansen panel cointegration test it has been found the existence of a long-run relationship between electricity consumption and per capita GDP for the panel of South Asian Countries as a whole. Thus VEC model has been considered to investigate the causal relationship between these two variables for the panel data. 4.3. Granger Causality

The cointegration relationship indicates that the existence of causal relationship between variables but it does not indicate the direction of causal relationship between variables. Therefore it is common to test for detecting the causal relationship between variables using the Engle and Granger test procedure. In the presence of cointegration relationship the application of Engle and Granger (1987) causality test in the first differenced variables by means of a VAR will misleading the results, therefore an inclusion of an additional variable to the VAR system such as the error correction term would help us to capture the long-run relationship. The augmented form of the Granger causality test involving the error correction term is formulated in a multivariate pth order vector error correction model.

pit 11k 12 it-k 1it1 1

it-1k=1it 21 22 it-k 2it2 2

LnEC LnECCECM

LnPGDP LnPGDPC

k

k k

β β ελ

β β ελ

∆ ∆ = + + + ∆ ∆

∑ (18)

where i = 1, 2,….,6; t = p+1, p+2,……….,T; .The C’s, 'sβ and 'sλ are the parameters to be estimated.,

it-1 ECM represents the one period lagged error-term derived from the cointegration vector and 'sε are

serially independent with mean zero and finite covariance matrix. From the equation (18) given the use of a VAR structure, variables are treated as endogenous variables. The F test is applied here to examine the direction of any causal relationship between the variables. The electricity consumption does not

Granger cause economic growth in the short run, if and only if all the coefficients 21kβ ’s ∀ k are not

significantly different from zero. Similarly the economic growth does not Granger cause electricity

consumption in the short run if and only if all the coefficients 12kβ ’s∀ k are not significantly different

from zero. There are referred to as the short-run Granger causality test. The coefficients on the ECM represent how fast deviations from the long-run equilibrium are eliminated. Another channel of causality can be studied by testing the significance of ECM’s. This test is referred to as the long run causality test. The panel short-run and long-run Granger causality results are reported in below in Table (6) and (7) Table 6: Result of the VAR and VEC Causality Analysis for Individuals

Country Model Lags Causality F-Test Prob LR Test (ECM)

Bangladesh VEC , Model 2 2 LnPGDP LnEC∆ → ∆ 0.1236 0.8841899 6.949897** LnEC LnPGDP∆ → ∆ 7.1709** 0.0030580 1.691778

India VEC, Model 2 2 LnPGDP LnEC∆ → ∆ 3.8283* 0.0339054 4.169677* LnEC LnPGDP∆ → ∆ 0.7586 0.4776890 1.931231

Iran VEC, Model 2 2 LnPGDP LnEC∆ → ∆ 0.1699 0.844593 8.848421**

LnEC LnPGDP∆ → ∆ 0.1238 0.884068 0.204216

Nepal VEC, Model 1 1 LnPGDP LnEC∆ → ∆ 5.2895* 0.028349 15.91295**

LnEC LnPGDP∆ → ∆ 1.1776 0.2862047 1.424880

Pakistan VAR 1 LnPGDP LnEC∆ → ∆ 3.8828* 0.0524859

LnEC LnPGDP∆ → ∆ 0.7007 0.4087578

Sri-Lanka VAR 1 LnPGDP LnEC∆ → ∆ 0.7412 0.3956965

LnEC LnPGDP∆ → ∆ 0.5012 0.4841119 x y → Means x Granger causes y. ** indicates significant at 1% level; * indicates significant at 5% level


325

Table 7: Panel Granger F-test results

Variables VEC ; [Lags 2], Model 1 VEC; [Lags 2 ] Model 2

LnEC∆ LnPGDP∆ ECM LnEC∆ LnPGDP∆ ECM

LnEC∆ 0.2465 1.240781 0.2539 1.194526

[0.7817524] [0.266671] [0.7760151] [0.27574]

LnPGDP∆ 0.0426 3.340317*** 0.0412 3.236719***

[0.95834] [0.06910672] [0.9596709] [0.073527] x y → Means x Granger causes y. The reported values in parentheses are the p-values of the test. *** indicates significant at10% level.

The findings in Table (6) indicate that there is short-run unidirectional causality running from

electricity consumption to economic growth in Bangladesh, from economic growth to electricity consumption in India, Nepal and Pakistan. There is no causal relationship between these variables in Iran and Sri-Lanka. It has also been found that there is a long-run relationship from Economic growth to electricity consumption for Bangladesh, India, Iran and Nepal out of six countries. The findings in Table (7) support that there is no short-run panel causal relationship between these two variables for South Asian countries but there is a long-run panel causal relationship running from electricity consumption to economic growth which is significant at 10% level of significance. 4.4. Short-Run and Long-Run Elasticity

The short run elasticity of economic growth with respect to electricity consumption can be obtained by estimating the following error correction model

it it it-1 itLnPGDP = LnEC + ECMα λ ε∆ ∆ + (19)

where itε is the random error terms, α and λ are the parameters to be estimated. For long-run

elasticity, now considering the following regression equation i i

i i

k

it i i it ij it-j ij it-j it

j=-p j=-k

LnPGDP = + LnEC + LnPGDP LnECp

uµ β λ γ∆ + ∆ +∑ ∑ (20)

This equation (21) is augmented with lead and lagged differences of the dependent and explanatory variables to control for serial correlation and endogeneous feedback effects. Here the GMM is applied to estimate both equation which control the problem of endegeneity and serial correlation of regressors. The estimated results are given below in Table (8) Table 8: Short-run and Long-run Elasticities for Individuals and also for Panel

Country

Short-run elasticity [ LnPGDP∆ is the

dependentvariable]

Long-run elasticity [ LnPGDP is the

dependent variable]

LnEC∆ ECM LnEC

Coefficient t-Test Coefficient t-Test Coefficient t-Test

Bangladesh 0.1430186** 2.58262 -0.237965** -2.99276 0.2663008** 15.14130 India 0.6036766** 5.99944 0.009850 0.17133 0.51734515** 10.35685 Iran 0.993714215 1.98529 -0.143507 -1.88532 -0.0311870 -0.24133 Nepal 0.26071572* 2.25001 -0.352105* -2.33445 0.4388449** 20.68710 Pakistan 0.3852596** 7.27048 -0.321399 -1.92713 0.4791014** 63.37425 Sri-Lanka 0.51074232** 10.45013 -0.642911** -3.97497 0.6124986** 97.05312

Panel 0.38127837** 4.77862 -0.0325689* -2.00313 0.54514213** 24.58911

**: Indicates significant at 1% level, *: indicates significant at 5% level.

The findings in Table (8) indicates that the short-run and long-run electricity consumption have

significant positive impact on economic growth for all the countries except Iran, but for Iran electricity consumption has insignificant negative impact on economic growth for long-run and insignificant


326

positive impact for short-run. The range of positive long-run elasticities of economic growth with respect to electricity consumption is from 0.6125 for Sri-Lanka to 0.2663 for Bangladesh and the range of short-run elasticities is from 0.9937 for Iran to 0.1430 for Bangladesh. For panel estimation it has been found that the electricity consumption has significant positive short-run and long-run impacts on economic growth. It has been found that the error correction term is statistically significant for Bangladesh, Nepal and Sri-Lanka represents evidence of a long run relationship between economic growth and electricity consumption of these countries. The error correction term for the panel is statistically significant, support the evidence of a long-run relationship between economic growth and electricity consumption for the panel of South Asian countries. Since the panel long-run elasticity of economic growth with respect to electricity consumption is (0.5451) higher than short-run elasticity (0.3813), indicates that over time higher electricity consumption in the panel of South Asian countries give rise to more economic growth.

5. Conclusions and Policy Implications This paper attempts to examine empirically the short-run and long-run causal relationship between economic growth and electricity consumption for the panel of South Asian countries using time series data for the period 1971- 2007. Also this study attempts to estimate the long-run and short-run elasticities of economic growth with respect to electricity consumption in order to examine the new approach which is proposed by Narayan and Narayan (2010). Before testing for any causal relationship between these variables within VAR structure at the first stage panel unit root tests and at the second stage panel cointegration analysis have been done. Four different panel unit root tests, Levin, Lin and Chu (LLC, 2002), Im, Peasaran and Shin (IPS, 2003), Maddala and Wu (1999), and Choi (2006) tests have been applied. These tests results support that the both variables are integrated of order one for all individual countries and also for the panel. The Johansen conintegration test results support that both the variables are cointegrated for Bangladesh, India, Iran and Nepal. Also the Johansen Fisher panel cointegration test results support that both the panel variables are cointegrated. Unidirectional short-run causal relationship has been found from electricity consumption to economic growth for Bangladesh, and from economic growth to electricity consumption for India, Nepal and Pakistan, no short-run causal relationship is found for Iran and Sri-Lanka. Also the test results support the evidence of long-run relationship from economic growth to electricity consumption for Bangladesh, India, Iran, and Nepal. The panel test results support that there is no short-run causal relationship between these two variables but there is an evidence of long-run panel causal relationship between the variables. It has been found that the short-run and long-run elasticities of economic growth with respect to electricity consumption are positively significant for all countries except Iran, but for Iran the long-run elasticity is negative but not significant. It has been found that the error correction term is statistically significant for Bangladesh, Nepal and Sri-Lanka represents evidence of a long run relationship between economic growth and electricity consumption of these countries. The error correction term for the panel is statistically significant represents evidence about a long-run relationship between economic growth and electricity consumption for the panel of South Asian countries. For the panel of South Asian countries it has been found that the long-run elasticity of economic growth with respect to energy consumption (0.5451) is higher than short run elasticity of 0.3812. This indicates that the economic condition will be in good in respect of electricity consumption. This means that over time higher electricity consumption in South Asian countries gives rise to more economic growth. Thus from the analytical results it can be concluded that any restriction on the use of electricity/energy in the region of Asia would be very harmful for economic development only for Bangladesh but other South Asian countries like India, Iran, Nepal, Pakistan and Sri-Lanka will not be affected by any policy of restriction on electricity/ energy use in the region of Asia.


327

References [1] Akarca,-A.T and Long,-I.T.V (1980) “On the Relationship between Energy And GNP: A

Reexamination”, Journal of Energy and Development, Vol. 5, pp. 326-331. [2] Asafu,-J. and Adjaye. (2000) “The Relationship between Energy Consumption, Energy Prices

and Economic Growth: Time Series Evidence from Asian Developing Countries”, Energy

Economics, Vol. 22, No. 6, pp. 615-625. [3] Cheng,-B.S and Lai,-T.W (1997) “An Investigation of Co-integration and Causality between

Energy Consumption and Economic Activity in Taiwan”, Energy Economics, Vol. 19, No. 4, pp. 435-444.

[4] Cheng,-B.S (1999) “Causality between Energy Consumption and Economic Growth in India: An Application of Cointegration and Error-Correction Modeling”, Indian Economic Review,

Vol. 34 , pp. 39–49. [5] Chang,-Y and Wong,-J.F (2001) “Poverty, Energy and Economic Growth in Singapore”,

Working Paper, Department of Economics, National University of Singapore, Vol. 37. [6] Choi,-I (2006) “Combination Unit Root Tests for Cross-sectionally Correlated Panels:

Econometric Theory and Practice”, Frontiers of Analysis and Applied Research: Essays in Honour of Peter C.B. Phillips. Cambridge University Press.

[7] Dirck,-C. B (2008) “Electricity Consumption and Economic Growth in the European Union: A Causality Study Using Panel Unit Root Test and Cointegration Analysis, IEEE Xplore.

[8] Erol,-U. and Yu,-E.S.H (1987) “On the Causal Relationship between Energy and Income for Industrialized Countries”, Journal of Energy and Development, Vol. 13, No.1 1, pp. 113-122.

[9] Imran,-K (2010) “Energy Consumption and Economic Growth: A Case Study of Three SAARC Countries”, European Journal of Social Sciences, Vol. 16, No. 2, pp. 206-224.

[10] Glasure,-Y.U and Lee,-A.R (1998), Cointegration, Error Correction and the Relationship between GDP and Energy: The case of South Korea and Singapore”, Resource and Energy

Economics, Vol. 20, No. 1, pp. 17-25. [11] Glasure,-Y.U (2002) “Energy and National Income in Korea: Further Evidence on the Role of

Omitted Variables”, Energy Economics, Vol. 24, pp. 355-365. [12] 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. [13] Kraft,-J. and Kraft,-A (1978) “On the Relationship between Energy and GNP”, Journal of

Energy and Development, 401-403. [14] Lee,-C.C (2005) “Energy Consumption and GDP in Developing Countries: A Cointegrated

Panel Analysis”, Energy Economics, Vol. 27, pp. 415-427. [15] Lee,-C.C (2006) “The Causality Relationship between Energy Consumption and GDP in G-11

Countries Revisited”, Energy Policy, Vol. 34, pp. 1086-1093. [16] Lee,-C.C and Chang,- C.P (2007) “Energy Consumption and Economic Growth in Asian

Economies: A More Comprehensive Analysis Using Panel Data”, Resource and Energy

Economics [17] 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, pp. 1-24. [18] Maddala,-G. S and Wu,-S (1999) “A Comparative Study of Unit Root Tests with Panel Data

and A New Simple Test”, Oxford Bulletin of Economics and Statistics, Vol. 61, pp. 631-652. [19] Masih,-A.M.M and Masih,-R. (1996) “Energy Consumption, Real Income and Temporal

Causality: Results from Multi-Country Study Based On Cointegration and Error-Correction Modeling Techniques”, Energy Economics, Vol. 18, pp., 165-183.

[20] Masih,-A.M.M and Masih,-R (1997) “On the Temporal Causal Relationship between Energy Consumption, Real Income, and Prices: Some New Evidence from Asian-Energy Dependent NICS Based on a Multivariate Cointegration/Vector Error-Correction Approach”, Journal of

Policy Modeling, Vol. 19, No. 4, pp. 417-440.


328

[21] Morimoto,-R and Hope,-C (2004) “The Impact of Electricity Supply On Economic Growth in Sri Lanka”, Energy Economics, Vol. 26, pp. 77-85.

[22] Narayan,-P. K and Singh,-B (2007) “The Electricity Consumption nnd GDP Nexus for The Fiji Islands”, Energy Economics, Vol. 29, pp. 1141-50.

[23] Narayan,-P.K and Russell,-S (2008) Energy Consumption and Real GDP in G7 Countries: New Evidence from Panel Cointegration with Structural Breaks”, Energy Economics, Vol. 30, pp. 2331–2341.

[24] Narayan,-P.K., Smyth,-R (2009) Multivariate Granger Causality between Electricity Consumption, Export and GDP: Evidence from A Panel of Middle Eastern Countries”, Energy

Policy, Vol. 37, pp. 229-36 [25] Narayan,-P.K., Narayan,-S and Popp S (2010) Does Electricity Consumption Panel Granger

Cause GDP? A New Global Evidence”, Applied Energy, Vol. 87, pp. 3294-3298 [26] Oh,-W. and Lee,-K (2004). Causal Relationship between Energy Consumption and GDP

Revisited: The Case of Korea 1970-1999”, Energy Economics, Vol. 26, pp. 51-59. [27] Oh,-W and Lee,-K (2004) “Energy Consumption and Economic Growth in Korea: Testing the

Causality Relation”, Journal of Policy Modeling, Vol. 26, pp. 973-981. [28] 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. 977-983. [29] Sari,-R and Soytas,-U (2007) The Growth of Income and Energy Consumption in Six

Developing Countries”, Energy Policy, Vol. 35, pp. 889–898. [30] Sari,-R and Soytas,-U (2009) “Are Global Warming and Economic Growth Compatible?

Evidence from Five OPEC Countries”, Applied Energy, Vol. 86, pp. 1887–1893. [31] Squalli,-J (2007) “Electricity Consumption and Economic Growth: Bounds and Causality

Analyses of OPEC Countries”, Energy Economics, Vol. 29, pp. 1192-1205. [32] Soytas,-U and Sari,-R (2003) Energy Consumption and GDP: Causality Relationship in G-7

Countries and Emerging Markets”, Energy Economics, Vol. 25, pp. 33-37. [33] Soytas,-U., Sari,-R. and Ewing,-B.T (2007) Energy Consumption, Economic Growth and

Carbon Emissions: Challenges Faced by an EU Candidate Member”, Ecological Economics, Vol. 68, No. 6, pp. 1667-1675.

[34] Soytas,-U., Sari,-R and Ewing, B.T (2007) Energy Consumption, Income, and Carbon Emissions in the United States”, Ecological Economics, Vol. 62, pp. 482–489.

[35] Stern,-D. I (1993) “Energy Growth in the USA: A Multivariate Approach”, Energy Economics, Vol. 15, pp. 137-150.

[36] Yang,-H.Y (2000), “A Note on the Causal Relationship between Energy and GDP in Taiwan,” Energy Economics, Vol. 22, No. 3, pp. 309-317.

[37] Yu,-E.S.H and Choi,- J.Y (1985) “The Causal Relationship between Energy and GNP: An International Comparison”, Journal of Energy and Development, Vol. 10, No. 2, pp. 249-272.

[38] Yu,-E.S.H and Hwang,-B.K (1984) “The Relationship between Energy and GNP: Further Results”, Energy Economics, Vol. 6, No. 3, pp. 186-190.

Copyright of European Journal of Social Science is the property of EuroJournals, Inc. and its content may not

be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

permission. However, users may print, download, or email articles for individual use.

Documents

67736459