Content uploaded by Thurai Murugan Nathan
Author content
All content in this area was uploaded by Thurai Murugan Nathan
Content may be subject to copyright.
© 2013 Research Academy of Social Sciences
http://www.rassweb.com 59
Journal of Empirical Economics
Vol. 1, No. 2, 2013, 5966
Does Electricity Consumption have Significant Impact towards the
Sectoral Growth of Cambodia? Evidence from Wald Test Causality
Relationship
Thurai Murugan Nathan1, Venus KhimSen Liew2
Abstract
It is recognized that energyoutput relationship studies from an economic perspective have become popular among
researchers recently. This study deals with the electricity consumption and outputs of some major economic sectors in
Cambodia. The newly developed ARDL bound testing approach has been employed to examine the cointegration
relationship. The Granger causality test of the aforementioned ARDL framework has also been used to investigate the
corresponding causality effect. Analysis and discussion based on the data of Cambodia, for the period of 1980 to 2010,
are presented. Interestingly, there is no any cointegration relationship identified between electricity consumption and
sectoral outputs. But, causality effect has been identified from electricity consumption towards the sectoral outputs.
Based on the findings, few recommendations have been made in the hope of systematizing the energy more efficiently
within the scrutinized sectors.
Key words: Sectoral Outputs, Electricity Consumption, ARDL Bound Testing, WALD test.
1. Introduction
The modelling of electricity consumption in every country is important since this source is getting expensive. The
energy crisis during the 1970s and high energy prices had negatively affected the economic performance of every sector
within the Asian region. In Cambodia, high GDP growth rate over the past 10 years has stimulated substantially
increasing demands for electricity within the country (Poch and Tuy, 2012). The recently published news from the
Business Monitor Online (2010) stated that the quality of electricity supply in Cambodia ranked at 121 out of 133
countries; the reason is that the country’s power supply services have been heavily damaged by war. Nevertheless, the
Cambodian government is recuperating under the support from the World Bank, ADB, Japan, USA, and European
Countries (General Directorate of Energy, MIME). At present, the electricity supply in Cambodia is fragmented into 24
isolated power systems centred in provincial towns and cities; all are fully reliant on diesel power stations. The per
capita consumption is only about 48 kWh/year and less than 15% of households have access to electricity (urban=53.6%
and rural=8.6%). The amount of electricity consumption in the private sector is 0.5%, the service sector takes up 40%,
and the industrial sector consumes 14% (World Bank, 2008).
MIME added that the main problem is that electricitysupplying companies such as Independent Power Producers
(IPP) and Electricite Du Cambogde (EDC) have distributed their cost energy into the smallest unit. This has made
energy expensive and the supply well below Cambodia's level of consumption3. In addition, since Cambodia is a
mediumincome developing country, being heavily dependent on imported energies can restrain the country’s economic
growth4. This problem is further aggravated when the energy sources are not managed properly. So, the main challenge
for the Cambodia government now is in meeting the rising electricity demand of a growing economy. Efficient
management of electricity usage can ensure economic growth consistency. Thus, it is crucial that considerable
importance is put into identifying the relationship between electricity consumption and sectoral outputs, since this
1 Universiti Tunku Abdul Rahman, Kampar, Perak, Malaysia.
2 Universiti Malaysia Sarawak, Kota Samarahan, Sarawak, Malaysia
3Poch and Tuy (2012) added that electricity cost in Cambodia remains one of the highest in the region and also the
world. Meanwhile, the electricity import rate from neighboring countries is increasing.
4 Cambodia mainly imports electricity from Thailand and Vietnam. The total electricity import in 2009 and 2010 are
842.40 million kWh and 1547.31 million kWh, respectively. The percentage change is recorded as 83.68% (EAC
Annual Report, 2010).
T. M. Nathan and V.K. Liew
60
approach can help to identify the electricity dependent sectors in Cambodia (Nwosa and Akinbobola, 2012; Liew et al.,
2012; Chebbi and Boujelbene, 2008)5.
The focus of this study is to identify how electricity consumption can affect a sector’s contribution in Cambodia
by using the ARDL bound testing and also the causality effect. The findings of this study will identify the energyled
sectors and sectors that are led by energy. If increased electricity consumption does not bring positive impact to the
output growth, then the sector is said to have inefficiently used the electricity supply; the electricity can thus be
allocated to other energydependent sectors to boost productivity. The critical part here is that this conservation policy
cannot be bluntly adopted since it will affect the production process and induce other side effects. A conservation
hypothesis has also been posed before this which states that policies implemented to reduce energy consumption may
have an adverse impact on the outputs. This has eventually become the driving force that motivates researchers to use
the causality test to identify the variables concerning a nation’s reliance on energy sources (AlIriani, 2006). To
conclude, it is anticipated that this study will become an asset to policy makers in managing the electricity consumption
of different sectors to overcome electricity shortage in Cambodia.
2. Methodology and Data
Data Description and model specification
Annual data from the year 1986 to 2010 were used in this study. These data on electricity consumption were
obtained from the U.S. Energy Information Administration (EIA) (2011). The data had been standardized using the
energy converter and were expressed in Million British Thermal Unit (Mmbtu). On the other hand, the data on sectoral
outputs for every sector were obtained from the United Nations (UN) (2011). The value added of the agricultural sector,
services sector, industrial sector, and transportation sector were used to represent sectoral outputs of the respective
sectors; this echoed with the method used by Liew et al., (2012) and Costantini and Martini (2010). All the variables
were changed into the logarithm form to reduce heteroskedasticity problems (Gujarati, 1995, pp.421).
The basic developed model for this study is as follows6. This model indicates that the total electricity consumption
is a function of sectoral growth.
=( ), (1)
The sectoral outputs under study were agricultural, industrial, services, and transportation outputs. In logarithmic
form, the model can be expressed as follows:
=1+2+, (2)
where = {, ,, }.
Unit Root Tests
The purpose of this unit root test is to determine whether the time series is consistent with the I(1) process with a
stochastic trend or with the I(0) process, which means it is stationary with a deterministic trend. When the number of
observation is low (shortspanned data), the unit root test will have little power (Chebbi and Boujelbene, 2008). To
overcome this problem, two types of unit root tests were used to examine the results, i.e., DickeyFuller GLS (DFGLS)
and PhillipsPerron (PP) unit root tests. According to Haug and Basher (2007), the DFGLS test generally has higher
power than the standard Augmented DickeyFuller (ADF). Shahbaz et al., (2009) reported that the commonly used ADF
test tends to overreject the null hypothesis of unit root, but the DFGLS has performed in this sense. The general
regression of the DFGLS test is:
= ++1+1
=1 +, (3)
where t is the time trend variable and k is the number of lags which are added to the model to ensure that the
residuals are white noise. The general regression for the PP test is:
=′+1+, (4)
5 The disaggregation data allowed comparisons to identify the impact of different energy sources on output rates (Sari et
al., 2008). In addition, a few studies such as Costantini and Martini (2010); Galindo (2005); Ruth and Lin (2006); and
Hunt et al., (2003) used sectoral outputs or production in their estimations.
6 The empirical model for this study is adopted from Loganathan et al., (2010).
Journal of Empirical Economics
61
where denotes the first difference, is the time series being tested, is a vector of deterministic terms
(constant, treand etc.), and is I(0) and may be heteroskedastic. Both tests have the same null hypothesis integrated to
the order of one, I(1), as stationary.
Cointegration test
Traditionally, if the variables were to be cointegrated, two conditions must be satisfied: the series for the
individual variables must have the same statistical properties, and the variables must be integrated in the same order. So,
if the variables have mixed results (integrated in I(0) and I(1)), the widely used Johansen analysis cannot be employed
in finding the cointegration relationship. To overcome this constraint, Pesaran and Shin (1999) and Pesaran et al.,
(2001) have developed the autoregressive distributed lag (ARDL) model or bound testing approach; this is the most
extensively used method in energy based studies7.
According to Wang (2009), the advantages of the ARDL model are that it can identify longrun relationships and
solve smallsample problems. The author further added that the unit root pretesting is not that critical when the ARDL
model is used because the integrated order, whether it is I(0) or I(1) or a mixture of both, is not important. Shraetha and
Chowdhury (2005) further added that this approach can take sufficient numbers of lags to capture the data generating
procedure in a generaltospecific modelling framework. The formulated ARDL model is as follows:
=+0,
1
=0
+1,
2
=1
+01+11+, (5)
=+0 ,
1
=0
+1,
2
=1
+01+11+, (6)
where is the firstdifference operator; 1 and 2 are the lag lengths; 0, and 1, (Eq. 5) and 0, and 1, (Eq. 6)
represent the shortrun dynamics of the model; 0 and 1 (Eq. 5) and 0 and 1 (Eq. 6) represent the longrun
relationship; and is a white noise error term. The null hypothesis in the ARDL is as follows:
0:0=1= 0 ;0 =1= 0, (7)
1:0 10 ;0 10. (8)
Besides that, this ARDL model also takes the error correction factors of previous periods into account. These error
correction terms, , and lag difference terms can test both shortterm and longterm relationships (Wang, 2009). The
general error correction model is as follows:
ln =+0,
1
=0
+1 ,
2
=1
+ +
=+ 0,
1
=0
+1,
2
=1
++
where and are the speeds of the adjustment parameter and is expected to be negative as well as statistically
significant. It indicates how fast the current differences in a dependent variable respond to the error correction term in
7Energy consumption can then be explained by its own lags and lagged value of a number of explanatory variables.
Dependent variables are also taken into account because it will help in future adjustments when the frequently
characterized energy consumption reacts to changes in the explanatory variables (Bentzen and Engsted, 2001).
T. M. Nathan and V.K. Liew
62
disequilibrium within the previous period. represents the residuals obtained from the estimated cointegration model
(Wang, 2009).
In this study, the cointegration relationship was examined using the Ftest in the ARDL framework. The computed
Ftest value was compared with the critical value in Table CI (iii) with unrestricted interception and no trend; tabulated
by Pesaran et al., (2001). In this ARDL framework, if the Ftest value is higher than the upper bound of the critical
value, then the null hypothesis will be rejected and it can be concluded that there is a longrun relationship among the
variables. If the Ftest value is less than the lower critical value, then the null hypothesis of the cointegration
relationship will be accepted, which means that no cointegration relationship exists among the variables. On the other
hand, if the Ftest value is between the lower and upper bound value, it means that the results are inconclusive (Pesaran
et al., 2001). However, there are a few steps that have to be taken care of before the cointegration test can be conducted.
Firstly, the optimum lag should be identified based on the SchwarzBayesian criterion (SBC) and/or Akaike Information
criterion (AIC). Then, the ordinary least square (OLS) technique should be used with the selected model to find out the
Ftest value. If this value is higher than the critical value, the null hypothesis of no longrun relationship among the
variables will be rejected. Finally, the longrun relationship and error correction model (ECM) can be estimated using
the selected lags. The ECM term should have a negative sign and should be statically significant.
Granger causality test
After the longrun relationship had been identified, the Granger Causality test was used to investigate the causal
relationships among the variables. This is because the cointegration test can only detect the longrun relationship, not
the direction of causality (Ozturk and Acaravci, 2010). According to Engle and Granger (1987), if the past value of is
able to predict the value, it is said that is the Granger cause of . Else, it can be said that is the Granger causes
of future values. The null hypothesis of the Granger causality test is that does not Granger cause and vice versa.
The following is the causality test model between the type of energy consumption and sectoral outputs:
=0+1
=1
+2
=1
++1
=0+1
=1
+2
=1
+2+2
where is the lag operator; α0, δ0, ,β’s and ø’ are the estimated coefficients; m and n are the optimal lags of the
series type of electricity consumptions and sectoral outputs (OUTPUT); ζ1t and ζ2t are serially uncorrelated random error
terms; μ1 and μ2 measure a single period responses of the variables to an exit to form equilibrium. is the error
correction term and the Ftest on the lagged explanatory variables represents the shortrun causal effect. If 2 is jointly
significant, then the null hypothesis which states that sectoral outputs does not Granger cause energy consumption can
be rejected (Odhiambo, 2009).
The Granger causality test based on ARDL framework (Wald test), known as the shortrun causality test, was used
in this study to find out the causality effect among the variables. The causality relationships were discovered by using a
common factor that had restricted the lags of the variables’ coefficient, which should be zero in this case. If the null
hypothesis of no causality can be rejected, it means that the variable are Granger caused (less than 5% of significance
level).
3. Empirical Evidence and Discussion
Table 1 summarizes the unit root test results of DFGLS and PP tests. For the DFGLS test, the agricultural and
electricity consumptions were integrated in I(1) and the rest were integrated in I(0). For the PP test, all variables were
integrated in I(1) except for the services outputs. This clearly indicates that there is a mixed order of I(0) and I(1) among
the variables.
Journal of Empirical Economics
63
Table 1: Unit Root Test Result
Variables
DickeyFuller GLS (ERS) test
PhillipsPerron (PP) test
Level
First Difference
Level
First Difference
Agricultural
2.202(0)
4.260(0)*
2.501(2)
4.944(1)*
Services
3.187(2)***

3.375(0)***

Industrial
4.139(5)*

1.959(0)
5.711(1)*
Transport
3.877(3)*

3.221(1)
6.186(3)*
Electricity
2.155(0)
3.087(0)*
2.266(2)
3.277(1)**
Notes: All variables had been transformed to natural logs. Asterisks (*), (**) and (***) indicate statistical
significant at the 1%, 5% and 10% levels, respectively. The optimum lag length for DFGLS test was autoselected
based on the Schwarz Info Criterion (SIC) and for the PP test, this was automatically selected based on the Newey
West Bandwidth.
Since the variables were integrated in a mixed order of I(0) and I(1), it is rational to adopt the ARDL model or
bound testing approach to investigate the cointegration relationship. Firstly, the bound testing process was executed to
obtain the Ftest values for all the combined variables. Then, the minimum SchwarzBayesian Criterion (SBC) value,
which was in accordance to Ozturk and Acaravci (2010) and Abdelhak et al.,’s (2011) work, was chosen as the
optimum lag in the process of developing the Fstatistic8. The chosen ARDL model must be superior in its statistical
properties so that its diagnostic tests will not demonstrate any substantiation of serial correlation, functional form,
normality, and heteroscedasticity (Verma, 2007)9. In this case, the chosen optimum lag was evaluated using the
diagnostic tests first. The Ftest value is accepted if it passes all diagnostic tests. Otherwise, another minimum SBC
value is chosen and then evaluated again until the best fitted model has been identified (Akindoade et al., 2008).
Table 2: Bound Testing Results
Dependent Variable
Independent Variable
Optimum Lag
FTest [Prob]
LNELEC
LNAGR
5
0.188 [0.834]
LNAGR
LNELEC
4
1.241 [0.339]
LNELEC
LNSER
4
2.551 [0.139]
LNSER
LNELEC
5
3.893 [0.096]
LNELEC
LNIND
5
1.613 [0.288]
LNIND
LNELEC
5
3.406 [0.117]
LNELEC
LNTRA
6
0.968 [0.508]
LNTRA
LNELEC
2
0.298 [0.747]
Notes: The critical values are; 1% (6.847.84), 5% (4.945.73) and 10% (4.044.78) significance level. The
optimum lag was selected using SBC. The maximum lag was fixed at 9.The critical values were obtained from
Table CI(iii) Case III: Unrestricted intercept and no trend reported in Pesaran et al., (2001).
Table 2 summarizes the bound testing results for all combined variables. The optimum lag chosen is the best fit
model for the combination where it passes all the diagnostic tests at 5% level of significance (Table 3). Surprisingly,
there was no evidence of cointegration relationship identified between the electricity consumption and sectoral outputs
in Cambodia since the obtained Ftest values were less than the critical value at 1%, 5% and 10% significance level.
Next, the Granger causality test was conducted to identify the causality effect among the combined variables.
Table 4 shows the shortrun causality test results obtained through the Wald test. The decision to reject the null
hypothesis is based on the probability value. If the probability value is less than that in the 5% significance level, the
null hypothesis which states that there is no causality relationship among the variables is rejected. In this study, it was
found that a unidirectional causality relationship running from electricity consumption towards industrial outputs in
shortrun existed. Consistent with the study of Altinay and Karagol (2005), an increase in electricity consumption can
8The lag selection process did not include. It will be provided upon request.
9Ibarra (2011) supported Verma (2007) statement by adding that “estimated equation must be confirmed by the battery
of diagnostic tests before moving on to the next step”.
T. M. Nathan and V.K. Liew
64
be viewed as the foremost indicator of industrial outputs growth in the country10. Such unidirectional causality
relationship was also found running from electricity consumption to agricultural and transport outputs; and at last
services from outputs to electricity consumption.
Table 3: Diagnostic Test Results
Dependent
Variable
Independent
Variable
Serial Correlation
a
[Prob]
Functional
Form b
[Prob]
Normality c
[Prob]
Heteroscedasticityd
[Prob]
LNELEC
LNAGR
0.435 [0.510]
0.643 [0.423]
0.208 [0.901]
0.604 [0.437]
LNAGR
LNELEC
0.571 [0.450]
0.044 [0.834]
0.714 [0.700]
0.071 [0.790]
LNELEC
LNSER
0.003 [0.954]
0.323 [0.570]
0.203 [0.903]
1.179 [0.278]
LNSER
LNELEC
1.939 [0.164]
7.396 [0.087]
1.036 [0.596]
0.437 [0.509]
LNELEC
LNIND
0.008 [0.930]
0.061 [0.805]
0.531 [0.767]
0.051 [0.822]
LNIND
LNELEC
2.851 [0.091]
7.772 [0.005]
2.182 [0.336]
0.309 [0.578]
LNELEC
LNTRA
0.616 [0.432]
0.695 [0.979]
0.403 [0.818]
0.028 [0.868]
LNTRA
LNELEC
0.430 [0.512]
3.555 [0.059]
2.802 [0.246]
0.106 [0.745]
Notes: aLagrange multiplier test of residual serial correlation. bRamsey's RESET test using the square of the fitted values.
cBased on a test of skewness and kurtosis of residuals. dBased on the regression of squared residuals on squared fitted values.
4. Conclusion and Recommendation
This study has applied the recently developed ARDL bound testing approach in identifying the cointegration
relationship and causality effect between the time series data of electricity consumption and sectoral growth in
Cambodia for the period covered from 1986 to 2009. The results of this study suggested that electricity consumption
does not give an impact in the longrun since there is no longrun relationship identified. Statistical evidences have also
showed that electricity consumption is an important component for sectoral growth in the shortrun in Cambodia
because of its causality effects that runs from electricity consumption towards the agricultural, industrial, and
transportation sectors. In here, the electricity consumption is the leading indicator of growing economy. In reality, the
electricity generation had increased from 0.2 billion kWh in 1995 to 1.5 billion kWh in 2009 (EIA, 2010) but it has yet
to meet the growing electricity demand in Cambodia. According to EIA (2010), the electricity consumption in the
production process had increased from 0.7 million kWh in 1995 to 5.3 million kWh in 2008; this corresponded to an
increase in sectoral inputs within the same time period11. In term of policy implication, the government of Cambodia
10Tang (2008) also used the same explanation as Altinay and Karagol (2005) in his study. Since electricity consumption
affects economic growth, he explained that the former is an important element for Malaysia’s economic growth and that
Malaysia is an electricitydependent country.
11The agricultural sector’s output increased from 1.32 billion in 1995 to 2.57 billion in 2008, that in the industrial sector
increased from 0.33 billion in 1995 to 1.78 billion in 2008, and the transport sector reported an increase in output from
0.15 billion in 1995 to 0.57 billion in 2008.
Table 4: Granger Causality Test Results for Cambodia (Wald Test Fstatistic)
Null Hypothesis
Wald Test
Direction
Chisquare
Prob
LNELEC does not Grangercause LNAGR
LNAGR does not Grangercause LNELEC
2.312
8.493*
0.128
0.004
ELEC AGR
LNELEC does not Grangercause LNSER
LNSER does not Grangercause LNELEC
2.414
9.079*
0.120
0.003
ELEC SER
LNELEC does not Grangercause LNIND
LNIND does not Grangercause LNELEC
6.692*
0.225
0.010
0.635
ELEC IND
LNELEC does not Grangercause LNTRA
LNTRA does not Grangercause LNELEC
4.830**
0.691
0.028
0.406
ELEC TRA
Notes: (**) and (*) denote the rejection of the null hypothesis at 5% and 1% significance level. Prob represents the
probability value.
Journal of Empirical Economics
65
should be more aware of the importance of stable electricity supply, and better coalition has to be formed with the
electricity companies in Cambodia. Policies that can boost the electricity supply should be drafted, researched, and
implemented.
Acknowledgement:
Research grant (FRGS/05(16)/729/2010(15) funded by the Malaysian Higher Ministry of Education is gratefully
acknowledged.
References
Abdelhak, S., Sulaiman, J. and Mohd, S., (2011). A Bound Testing to Cointegration: An Examination of Natural
Disasters and GDP Relationship in Southern Africa Region, the International Journal of Applied Economics
and Finance, 5(3), 213225.
Akinboade, O. A., Ziramba, E. and Kumo, W. L., (2008). The Demand for Gasoline in South Africa: An Empirical
Analysis Using CoIntegration Techniques, Energy Economics, 30, 32223229.
AlIriani, M. A., (2006). EnergyGDP relationship revisited: an example from GCC countries using panel causality.
Energy Policy, 34, 33423350.
Altinay, G. and Karagol, E., (2005). Electricity Consumption and Economic Growth: Evidence from Turkey, Energy
Economics, 27, 849856.
Bentzen, J. and Engsted, T., (2001). A Revival of the Autoregressive Distributed Lag Model in Estimating Energy
Demand Relationships. Energy Economics, 26(1), 45–55.
Chebbi, H. E., and Boujelbene, Y., (2008). Agricultural and NonAgricultural Outputs and Energy Consumption in
Tunisia: Empirical Evidences from CoIntegration and Causality. 12th Congress of the European Association
of Agricultural Economists – EAAE 2008.
Costantini, V., and Martini, C., (2010). The Causality between Energy Consumption and Economic Growth: A Multi
Sectoral Analysis Using NonStationary Cointegration Panel Data. Energy Economics, 32, 591603.
Engle, R. F., and Granger, C. W. J., 1987. Cointegration and Errorcorrection: Representation, Estimation and Testing.
Econometrica, 55, 251–276.
Galindo, L. M., (2005). Short and LongRun Demand for Energy in Mexico: A Cointegration Approach. Energy
Policy, 33, 11791185.
Gujarati, D., N., 1995. Basic Econometrics. 4th ed. McgrawHill, New York, ISBN 0070252149.
Haug, A. A., and Baster, S. A., (2007). Linear or Nonlinear cointegration in the purchasing power parity relationship?
(Economics Discussion Papers), 0712. Retrieved from Department of Economics, University of Otago.
Available from: http://www.business.otago.ac.nz/econ/ research/discussionpapers/ index.html.
Hunt, L. C., Judge, G., and Ninomuya, Y., (2003). Underlying Trend and Seasonality in UK Energy Demand: A
Sectoral Analysis. Energy Economics, 25, 93118.
Ibarra, C. A., (2011). Import Elasticities and the External Constraint in Mexico, Economic Systems, 35(3), 363–377.
Liew, V. K. S., Nathan, T. M. and Wong, W. K., (2012). Are Sectoral Outputs in Pakistan Led by Energy
Consumption?, Economics Bulletin, 32(3), 23262331.
Loganathan, Nanthakumar and Subramaniam, T., (2010). Dynamic Cointegration Link between Energy Consumption
and Economic Performance: Empirical Evidence from Malaysia. International Journal of Trade, Economics
and Finance, 1(3), 023.
Odhiambo, N. M., (2009). Electricity Consumption and Economic Growth in South Africa: A Trivariate Causality Test,
Energy Economics, 31(5), 635640.
Ozturk, I., and Acaravci, A., (2010). The Causal Relationship between Energy Consumption and GDP In Albania,
Bulgaria, Hungary and Romania: Evidence from ARDL Bound Testing Approach, Applied Energy, 87, 1938
1943.
T. M. Nathan and V.K. Liew
66
Pesaran, M. H. and Shin, Y., (1999). An Autoregressive Distributed Lag Modeling Approach to Cointegration Analysis,
Chapter 11 in Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial
Symposium, Strom S (Ed.), Cambridge University Press: Cambridge.
Pesaran, M. H., Shin, Y. and Smith, R.(2001) Bounds Testing Approaches to the Analysis of Level Relationships,
Journalof Applied Econometrics, 16,289 – 326.
Poch, K. and S. Tuy (2012), Cambodia’s Electricity Sector in the Context of Regional Electricity Market Integration’ in
Wu, Y., X. Shi, and F. Kimura (eds.), Energy Market Integration in East Asia: Theories, Electricity Sector and
Subsidies, ERIA Research Project Report 201117, Jakarta: ERIA, pp.141172.
Ruth, M., and Lin, A., (2006). Regional Energy Demand and Adaptations to Climate Change: Methodology and
Application to the State of Maryland, USA. Energy Policy, 34, 28202833.
Sari, R., Ewing, B. T., and Soytas, U., (2008). The Relationship between Disaggregate Energy Consumption and
Industrial Production in the United States: An ARDL Approach. Energy Economics, 30, 23022313.
Shahbaz, M., Awan, R. U., and Nasir, N. M., (2009). Producer and Consumer Prices Nexus: ARDL Bounds Testing
Approach. International Journal of Marketing Studies, 1(2), 7886.
Shrestha, M. B., and Chowdhury, K., (2005). ARDL Modeling Approach to Testing the Financial Liberalization
Hypothesis, University of Wollongong, Economics Working Paper Series 2005, 0515.
Tang, C. F., (2008). A ReExamination of the Relationship between Electricity Consumption and Economic Growth in
Malaysia. Energy Policy, 36, 30773085.
United Nation (UN). [Online]Available at: <http://www.unstats.un.org/unsd/snaama/ dnltransfer.asp?fid=16> [Accessed
on 10 May 2011].
U.S. Energy Information Administration (EIA). Retrieved from: http://205.254.135.7
/cfapps/ipdbproject/IEDIndex3.cfm?tid=5andpid=5andaid=2. Accessed on 15 May 2011.
U.S. Energy Information Administration (EIA), 2011.[Online] Available at: <http:// www.eia.gov/countries/country
data.cfm?fips=bm>[Accessed on 20 April 2012].
Varma, R., (2007). Saving Investment and Growth in India: An Application of the ARDL Bounds Testing Approach,
South Asia Economic Journal, 8(1), 8798.
Wang, Y. S., (2009). The Impact of Crisis Events and Macroeconomic Activity on Taiwan’s International Inbound
Tourism Demand, Tourism Management, 30, 758.