CO2 emissions, energy consumption and economic growth in Association of Southeast Asian Nations (ASEAN) countries: A cointegration approach Behnaz Saboori, Jamalludin Sulaiman Economic Programme, School of Social Sciences, Universiti Sains Malaysia, 11800 USM Penang, Malaysia a r t i c l e i n f o Article history: Received 28 November 2012 Received in revised form 15 February 2013 Accepted 14 April 2013 Available online 20 May 2013 Keywords: Carbon dioxide emissions Energy consumption Economic growth a b s t r a c t This study examines the cointegration and causal relationship between economic growth, carbon dioxide (CO2) emissions and energy consumption in selected Association of Southeast Asian Nations (ASEAN) countries for the period 1971e2009. The recently developed Autoregressive Distributed Lag (ARDL) methodology and Granger causality test based on Vector ErrorCorrection Model (VECM) were used to conduct the analysis. There was cointegration relationship between variables in all the countries under the study with statistically significant positive relationship between carbon emissions and energy consumption in both the short and longrun. The longrun elasticities of energy consumption with respect to carbon emissions are higher than the shortrun elasticities. This implies that carbon emissions level is found to increase in respect to energy consumption over time in the selected ASEAN countries. A significant nonlinear relationship between carbon emissions and economic growth was supported in Singapore and Thailand for the longrun which supports the Environmental Kuznets Curve (EKC) hypothesis. The Granger causality results suggested a bidirectional Granger causality between energy consumption and CO2 emissions in all the five ASEAN countries. This implies that carbon emissions and energy consumption are highly interrelated to each other. All the variables are found to be stable suggesting that all the estimated models are stable over the study period.
Trang 1See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/257177334
CO2 emissions, energy consumption and economic growth in Association of Southeast Asian Nations (ASEAN) countries: A cointegration approach
Article in Energy · June 2013
DOI: 10.1016/j.energy.2013.04.038
CITATIONS
130
READS
1,132
2 authors, including:
Some of the authors of this publication are also working on these related projects:
Asia Pacific gas market View project
Behnaz Saboori
Islamic Azad University, South Tehran Branch
25 PUBLICATIONS 874 CITATIONS
SEE PROFILE
All content following this page was uploaded by Behnaz Saboori on 10 April 2018.
The user has requested enhancement of the downloaded file.
Trang 2CO 2 emissions, energy consumption and economic growth in
Association of Southeast Asian Nations (ASEAN) countries:
A cointegration approach
Economic Programme, School of Social Sciences, Universiti Sains Malaysia, 11800 USM Penang, Malaysia
a r t i c l e i n f o
Article history:
Received 28 November 2012
Received in revised form
15 February 2013
Accepted 14 April 2013
Available online 20 May 2013
Keywords:
Carbon dioxide emissions
Energy consumption
Economic growth
a b s t r a c t
This study examines the cointegration and causal relationship between economic growth, carbon dioxide (CO2) emissions and energy consumption in selected Association of Southeast Asian Nations (ASEAN) countries for the period 1971e2009 The recently developed Autoregressive Distributed Lag (ARDL) methodology and Granger causality test based on Vector Error-Correction Model (VECM) were used to conduct the analysis There was cointegration relationship between variables in all the countries under the study with statistically significant positive relationship between carbon emissions and energy con-sumption in both the short and long-run The long-run elasticities of energy concon-sumption with respect to carbon emissions are higher than the short-run elasticities This implies that carbon emissions level is found to increase in respect to energy consumption over time in the selected ASEAN countries A sig-nificant non-linear relationship between carbon emissions and economic growth was supported in Singapore and Thailand for the long-run which supports the Environmental Kuznets Curve (EKC) hy-pothesis The Granger causality results suggested a bi-directional Granger causality between energy consumption and CO2emissions in all thefive ASEAN countries This implies that carbon emissions and energy consumption are highly interrelated to each other All the variables are found to be stable sug-gesting that all the estimated models are stable over the study period
Ó 2013 Elsevier Ltd All rights reserved
1 Introduction
The debate about the relationship between energy consumption
and economic development stems from the increasing effects of
energy on economic development With growing concerns about
global warming or climate change, there is a pressure for nations to
consume a balanced level of energy that control the emissions to
the environment; but at the same time ensuring the country’s
sustainable economic growth
Testing the relationship between economic growth and
envi-ronmental pollution under the Envienvi-ronmental Kuznets Curve (EKC)
hypothesis forms the first group of related literatures The EKC
hypothesis claims an inverted U-shaped relationship between
environmental pollution and income per capita However related
empirical studies are inconclusive Although Shafik and
Bandyo-padhyay[1], Seldon and Song[2], Unruh and Moomaw[3], Galeotti
and Lanza[4], Dinda and Coondoo[5]and Managi and Jena[6]have
found evidence supporting the existence of an EKC, many others have found evidences against the hypothesis (see e.g Refs.[7e12]).1 The second group of studies examined the relationship between economic growth and energy consumption Since the seminal study by Kraft and Kraft [16], more literatures have emerged examining the cointegration and causality relationship between economic growth and energy consumption (see Refs.[17e20]for recent studies)
Later due to the omitted variable biased embodied in thefirst and second group of literatures, a third group was formed, which tested the relationship between economic growth, energy con-sumption and pollution emissions in a multivariate framework They address the problem of omitted variable bias that may yield spurious results for the EKC hypothesis and the nature of the causality
Some recent examples of studies for developed countries include by Ang[21] for France, Soytas et al.[22] for the United
* Corresponding author Tel.: þ60 176168939.
E-mail address: behnazsaboori@yahoo.com (B Saboori).
1 An extensive review of EKC literature is available in papers of Stern [13] , Dinda
[14] and Kijima et al [15]
Contents lists available atSciVerse ScienceDirect
Energy
j o u r n a l h o m e p a g e : w w w e l s e v i e r c o m / l o c a t e / e n e r g y
0360-5442/$ e see front matter Ó 2013 Elsevier Ltd All rights reserved.
Energy 55 (2013) 813e822
Trang 3States and Acaravci and Ozturk[23]for nineteen European
coun-tries There are recent studies for developing countries such as Ang
[24]for Malaysia, Zhang and Cheng[25]for China, Ghosh[26]for
India, Lotfalipour et al.[27]for Iran, Menyah and Rufael[28]for
South Africa and Ozturk and Acaravci[29]for Turkey.2However no
consensusfinding has emerged from these studies This makes the
recommendation of a unique policy across countries impossible at
this point in time
For example, Ang[21]argues an inverted U-shaped relationship
between CO2emissions and output for France thus suggesting the
evidence of EKC He found a long-run relationship between output,
CO2emissions and energy consumption with a causal relationship
from output to energy consumption and CO2emissions in the
long-run and from energy consumption to economic growth in the
short-run Soytas et al.[22]found that income does not cause CO2
emissions in the United States in the long-run, but energy
con-sumption does Using Autoregressive Distributed Lag (ARDL)
bounds testing approach of cointegration for nineteen European
countries Acaravci and Ozturk[23]found a cointegration
relation-ship between CO2emissions, energy consumption and real income
for Denmark, Germany, Greece, Iceland, Italy, Portugal and
Switzerland The EKC was satisfied just in the cases of Denmark and
Italy Ang[24]and Zhang and Cheng[25]in the cases of Malaysia
and China respectively found a unidirectional Granger causality
running from GDP to energy consumption in long-run
Ghosh[26]showed that there is a bi-directional causality
be-tween carbon emissions and economic growth in India and a
uni-directional causality running from economic growth to energy
supply and energy supply to carbon emissions in the short-run
Contrary to other studies, Menyah and Rufael[28]in a study for
South Africa found a unidirectional causality running from pollutant
emissions to economic growth and from energy consumption to
economic growth and CO2 emissions Similarly, Lotfalipour et al
[27]found a unidirectional causality from energy consumption to
CO2emissions for Iran Ozturk and Acaravci[29]show that neither carbon emissions per capita nor energy consumption per capita
influences real GDP (Gross Domestic Product) per capita in Turkey Luzzati and Orsini[36]investigated the relationship between ab-solute energy consumption and GDP per capita for 113 countries The results did not support an energy-EKC hypothesis for the world
as a single unit however, they found a positive monotonic rela-tionship between carbon emissions and economic growth
In 1967, the Association of Southeast Asian Nations (ASEAN) was formed consisting of Indonesia, Malaysia, Philippines, Singapore and Thailand Since then membership has expanded to include Brunei Darussalam, Vietnam, Laos, Myanmar and Cambodia mak-ing up what is today the ten member states of ASEAN The region is surrounded by major seas and gulfs such as the South China Sea, the Andaman Sea and the gulf of Thailand It has a total land area of 4.436 million square kilometers (3.3% of the world’s land area) and
a total population of 584 million (8.7% of the world population) ASEAN is one of the fastest growing economic regions in the world Its economy has experienced a rapid GDP growth at an average annual rate of 4.8 and 6.5 percent for 1994e1999 and
2000e2008 periods respectively Continuous growth in urbaniza-tion and industrializaurbaniza-tion in the region increase energy consump-tion substantially With the assumed GDP and populaconsump-tion growth rate, thefinal energy consumption is estimated to increase at an average annual rate of 4.4 percent in 2030[37] This growth is very much higher than the world’s average growth rate of 1.4 percent per year in energy demand over 2008e2035[38] In addition CO2
emissions are increasing in a similar way.Fig 1shows that carbon emissions, energy consumption and economic growth are rapidly increasing in ASEAN countries over the period 1971e2008 Thus it
is justifiable to investigate the long-run relationship and causality issues between the variables for these countries
Surprisingly despite of the importance of the region, there has been no published empirical study examining the relationship between environmental pollution, economic growth and energy consumption for each of the ASEAN countries This study employs a time series analysis of cointegration and causal relationship be-tween economic growth, energy consumption and CO emissions
8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5
CO emissions per capita in logs
8.4 8.8 9.2 9.6 10.0
Energy consumption per capita in logs
8.8 9.2 9.6 10.0 10.4 10.8 11.2
GDP per capita in logs
Fig 1 Time series plots of real GDP per capita (constant US$), carbon emissions per capita (metric tons oil equivalent) and per capita energy consumption (kg of oil equivalent) in log levels in the ASEAN countries.
2 There are panel-based Granger causality studies of the CO 2 emissions-economic
growth-energy consumption nexus such as Apergis and Payne [30] for Central
American countries, Apergis and Payne [31] for a panel of the Commonwealth of
Independent States, Lean and Smyth [32] for five members of ASEAN countries, Pao
and Tsai [33] for BRIC (Brazil, Russia, India and China) countries, Al-mulali and Binti
Che Sab [34] for Sub Saharan African countries and Bashiri Behmiri and Pires Manso
[35] for OECD countries.
B Saboori, J Sulaiman / Energy 55 (2013) 813e822 814
Trang 4for initialfive ASEAN countries (Indonesia, Malaysia, Philippines,
Singapore and Thailand) over the period 1971e2008 The recently
developed ARDL bounds testing approach of cointegration by
Pesaran and Shin[39] and Pesaran et al [40]and Vector
Error-Correction Model (VECM) based Granger causality tests were used
The rest of the paper is organized as follows The next section
presents the methodology and data The third section reports the
empirical results and the last section concludes the paper
2 Methodology
Following the methodology of recent studies by Halicioglu[41],
Sari and Soytas [42], Menyah and Rufael [28] and Ozturk and
Acaravci [29], this study employs the ARDL approach to
cointe-gration test developed by Pesaran and Shin[39]and Pesaran et al
[40]and the VECM based Granger causality method, to investigate
the long-run and the causal relationship between economic
growth, CO2emissions and energy consumption for thefive ASEAN
countries during the period 1971e2008
2.1 Bounds testing approach to cointegration
Testing for the existence of cointegration among variables is
important The existence of cointegration among variables not
only shows a long-run equilibrium relationship between
vari-ables but also it can guarantee consistent results when the
or-dinary least square (OLS) method is used for estimation of the
coefficients
ARDL, a relatively new cointegration technique which has been
introduced by Pesaran et al.[40], has many advantages over other
cointegration approaches The ARDL approach does not require
establishing the order of integration of the variables (unit-root
test) The approach is applicable regardless of whether the
under-lying regressors are I(0), I(1) or fractionally integrated Other
standard cointegration approaches such as EngleeGranger[43]and
JohanseneJuselius [44] require that variables be integrated at
unique level of integration The ARDL approach is thus free of
pre-testing problems associated with the order of integration of
vari-ables Second the short-run as well as the long-run effects of the
independent variables on the dependent variable are assessed
simultaneously, which allows researchers to distinguish between
the short-run and long-run effects of the variables Third, the ARDL
approach has better properties for small samples Pesaran and Shin
[39]showed that with the ARDL framework, the OLS estimators of
the short-run parameters are consistent and the ARDL based
esti-mators of the long-run coefficients are super consistent in small
sample sizes Finally, all variables are assumed to be endogenous so
the endogeneity problems associated with the EngleeGranger
method are avoided
In order to establish the relationship between CO2 emissions,
economic growth and energy consumption for each of the selected
five ASEAN countries the following model is proposed
LEt ¼ b0þb1LYtþb2ðLYtÞ2þb3LENtþ 3t (1)
where E is per capita CO2emissions, Y represents per capita real
income, EN stands for energy use per capita and 3tis the standard
error term Based on EKC hypothesis, the sign ofb1is expected to
be positive, whereas a negative sign is expected forb2 Since a
higher level of energy consumption leads to greater economic
activity and stimulates CO2 emissions, b3 is expected to be
positive
Equation(1)shows the long-run relationships among the
un-derlying variables To implement the ARDL approach to
cointegra-tion into these model, first the short-run dynamics need to be
added into the long-run The short-run equation corresponding to the long-run equations of (1) is written as equation(2)
k ¼ 1
k ¼ 0
k ¼ 0
þXn
k ¼ 0
þf4LENt1þ 3t
(2)
First, equation(1) is estimated by the OLS method Then the F-statistic for joint significance of the variables needs to be calcu-lated The null hypotheses for this test for equation(2)is as follows,
H0 ¼f1 ¼f2 ¼ f3 ¼f4 ¼ 0 which are tested against its alter-native H1 ¼f1sf2sf3sf4s0 The F-test is conducted to test the existence of a long-run relationship among the variables The critical values of the F-statistics in this test are available in Pesaran and Pesaran [45]and Pesaran et al.[40] However, these critical values are generated for sample sizes of 500 and 1000 observations Narayan[46]argues that exiting critical values cannot be used for small sample sizes because these values were obtained based on large sample sizes Narayan [46] calculated critical values for sample sizes ranging from 30 to 80 observations Given the small sample size in this study which is only 39, the critical values of Narayan[46]for the bounds F-test are employed
There are two sets of critical values for a given significance level, with and without a time trend, one for I(0) variables and the other set for I(1) variables, which are known as lower bounds (LCB) and upper bounds critical values (UCB) respectively This provides a band covering all possible classifications of the variables into I(0) and I(1) If the computed F-statistic is higher than the UCB, the null hypothesis of no cointegration is rejected and if it is below the LCB the null hypothesis cannot be rejected, and if it lies between the LCB and UCB, the result is inconclusive At this stage of the estimation process, the unit-root tests are normally carried out on variables entered into the model
The modified ARDL approach estimates (p þ 1)k number of regression in order to obtain optimal lag length for each variable, where‘p’ is the maximum number of lags to be used and “k” is the number of variables in the model Bahmani-Oskooee and Goswami
[47]indicate that the F-statistic is affected by the number of lags entered into the model Therefore, there is a need to choose the appropriate number of lags in the model The lag orders of the variables can be selected on the basis of R2, SchawrtzeBayesian criteria (SBC), HannaneQuinn Criterion (HQC) and Akaike’s infor-mation criteria (AIC) The SBC selects the smallest possible lag length while AIC is employed to select maximum relevant lag length The long-run relationship among variables can be estimated after the selection of the ARDL model by AIC or SBC criterion
An Alternative way to test for the existence of a long-run rela-tionship among the variables of the model is to substitute the lagged level variables with an error-correction term (ECT) and test for the significance of its coefficient To obtain these coefficients, short-run error-correction equation in (2) need to be estimated Then the ECT can be calculated as the sum of lagged level terms using the estimates off1 In the next step, the lagged level term in each equation will be replaced by the lagged value of constructed ECT and the model is estimated one more time with the same op-timum number of lags selected by AIC or SBC The ECT indicates the speed of the adjustment and shows how quickly the variables re-turn to the long-run equilibrium and it should have a statistically significant coefficient with a negative sign, then the cointegration relationship exists The general ECM (Error Correction Model) of equation(2)is formulated as equation(3)
Trang 5DLEt ¼ d0þXn
k ¼ 1
k ¼ 1
k ¼ 1
þXn
k ¼ 1
(3)
After testing for the existence of the long-run relationship
among the variables in the model, one can proceed to the next stage
and estimate the long-run relations in equation(1)
To ensure the goodness of fit of the model, the diagnostic
and stability tests are also conducted These include, testing
for serial correlation, functional form, normality and
hetero-scedasticityassociated with selected model Furthermore
Pesaran et al [40] suggested estimating the stability of long
and short-run estimate through cumulative sum (CUSUM) and
cumulative sum of squares (CUSUMSQ) tests proposed by
Brown et al [48] In order to check the stability of the
long-run and the short-long-run coefficients CUSUM and CUSUMSQ are
employed Graphically, these two statistics are plotted within
two straight lines bounded by the 5% significance level If any
point lies beyond this 5% level, the null hypothesis of stable
parameters is rejected
2.2 Granger causality test
The cointegration relationship between CO2 emissions,
eco-nomic growth and energy consumption is investigated with the
use of ARDL bounds testing approach, but it does not indicate
the direction of causality between variables Identifying the
causal direction between CO2emissions, economic growth and
energy consumption provides policy makers with a clearer
un-derstanding of the role of energy consumption constraints on
CO2 emissions and economic growth This paper employs
Granger causality test based on VECM to examine the causal
relationship between mentioned variables The Engle and
Granger [43] causality test in the first difference variable by
means of a VAR (Vector Autoregressive) model will give
misleading results in the presence of cointegration Therefore it
is necessary to include the Error-Correction Term (ECT) as an
additional variable to the VAR system The direction of causality
can be detected through the VECM of long-run cointegration
The augmented form of Granger causality test with ECM is formulated in multivariaterth order of VECM model as follows:
ð1 BÞ
2 6 6
LEt
LYt
LY2 t LENt
3 7
7 ¼
2 6 6
c1
c2
c3
c4
3 7
7þX p i¼ 1 ð1 BÞ
2 6 6
d11;id12;id13;id14;i
d21;id22;id23;id24;i
d31;id32;id33;id34;i
d41;id42;id43;id44;i
3 7 7
2 6 6
LEti
LYti
LYti2 LENti
3 7
7þ
2 6 6
l1
l2
l3
l4
3 7 7½ECt1 þ
2 6 6
g1t
g2t
g3t
g4t
3 7 7 (4)
where (1 B) is the lag operator and ECt 1is error-correction term.
Residual terms,gt’s are uncorrelated random disturbance term with zero mean and d’s are parameters to be estimated The significant t-statistics on the coefficients of the lagged ECTs indicate the signif-icance of the long-run causal relationships, while F-statistic or Wald test investigate short-run causality through the significance
of the lagged independent variables The AIC and SBC criteria were used to choose the appropriate lag length
2.3 Data
Data for Indonesia, Malaysia, Philippines, Singapore and Thailand for the period of 1971e2009 was chosen on the basis of their availability Other ASEAN member countries do not have a complete set of all the series and thus not selected for the study Real GDP per capita (Y) in constant 2000 US$, CO2emissions (E)
in metric tons per capita and per capita energy consumption (EN)
in kg of oil equivalent were used CO2emissions are those which stemming from the burning of fossil fuels and the manufacture of cement They include carbon dioxide produced during con-sumption of solid, liquid, and gas fuels and gasflaring Energy use refers to use of primary energy before transformation to other end-use fuels, which is equal to indigenous production plus imports and stock changes, minus exports and fuels supplied to ships and aircraft engaged in international transport All data are from World Development Indicators (WDI) online database
Table 1gives the summary statistics of each of the variable used
in the analysis
Table 1
Descriptive statistics for the five selected ASEAN countries.
Descriptive
statistics
E indicates per capita carbon dioxide emissions in metric tons, Y indicates per capita real GDP in constant 2000 US$ and EN indicates per capita energy consumption in kg of oil equivalent.
B Saboori, J Sulaiman / Energy 55 (2013) 813e822 816
Trang 63 Thefindings
The augmented Dickey and Fuller[49]and Phillips and Perron
[50] tests were used to identify the order of integration of the
variables.3In both tests the null hypothesis of the series has a unit
root is tested against the alternative of stationarity.Table 2
sum-marizes the outcome of the ADF (Augmented DickeyeFuller) and
PP (PhillipsePerron) unit-root tests on the natural logarithms of the
levels and thefirst differences of the variables The results suggest
that all series are stationary in theirfirst differences, indicating that
none of the variable is I(2) or beyond Hence validate the use of
bounds testing for cointegration
The ARDL bounds testing approach starts with the F-test to
confirm the existence of cointegration between the variables in
equation(2) The maximum lags are selected after applying several
misspecification tests to ensure that the classical regression
as-sumptions are not violated The optimum lags are selected relying
on minimizing the AIC The maximum lag order 5, 3, 2, 4 and 3 were
set for Indonesia, Malaysia, Philippines, Thailand and Singapore respectively With that maximum lag lengths setting, the ARDL (w,
x, y, z) models are selected using AIC.4 The results of cointegration inTable 3show that the F-statistic is greater than its upper bound critical value (3.898 at 10%) for Singapore and Thailand, so the evidence of cointegration While in other cases cointegration is supported by the significantly negative coefficient obtained for ECt1.5This term shows the speed of the adjustment process to restore equilibrium The relatively high ECt1 coefficients imply a faster adjustment process The values of the coefficients of ECt 1in most of the cases are quite high, indicating
the high speed of adjustment to the long-run equilibrium following short-run shocks
Table 4 presents the long-run estimation results along with diagnostic tests such as serial correlation, functional form, normality and heteroscedasticity The significant positive and negative coefficients of LY and (LY)2with respect to environmental emissions provide evidence of EKC This suggests that carbon
Table 2
Unit-root tests results.
Note: 1 ***, ** and * are 1%, 5% and 10% of significant levels, respectively 2 The optimal lag length was selected automatically using the Schwarz information criteria for ADF test and the bandwidth is selected using the NeweyeWest method for PP test E indicates per capita carbon dioxide emissions in metric tons per capita, Y indicates per capita real GDP in constant 2000 US$ and EN indicates per capita energy consumption in kg of oil equivalent.
3 ARDL bounds testing approach is applicable for the variables that are I(0) or I(1)
and in the presence of I(2) variables, the computed F-statistics provided by Pesaran
4 ARDL (w, x, y, z) represents the ARDL model in which the variables take the lag length w, x, y and z, respectively.
5
Trang 7emissions per capita increases with the increase of economic
growth but after a certain level of GDP which is the turning point, it
starts to decrease
There is significant positive and negative coefficients of LY and
(LY)2 with respect to environmental emissions in the cases of
Singapore and Thailand thus provide evidence of EKC The long-run
elasticity of carbon dioxide emissions per capita with respect
to real GDP per capita is 7.8326 0.905LY for Singapore and
9.0158 1.2071LY for Thailand The turning point of per capita real
income is Y*¼ b1/2b2.6Based on these results the turning points
are calculated in logarithms The turning point of per capita real
income turned out to be 8.65, compared to the highest value of LY
for Singapore which is 10.31 In the case of Thailand, the turning
point of per capita real income turned out to be 7.47, compared to
the highest value of LY for Thailand which is 7.87
Results indicated negative and positive coefficients at 10%
significance level for GDP and square of per capita real GDP
respectively for Philippines In the case of Indonesia, the results
also revealed negative coefficient for real GDP per capita at 5%
significance level and positive coefficient for square of per capita
real GDP at 1% significance level Based on the empirical findings,
Indonesia and Philippines are currently on the increasing part of
the EKC curve These results do not support the EKC in Indonesia,
Malaysia and Philippines under our long-run analysis These
findings are in line with findings of Ozturk and Acaravci[29]who
did notfind EKC hypothesis at causal framework by using a linear
logarithmic model in the case of Turkey Furthermore, Akbostanci
et al.[12]found a monotonically increasing relationship between
CO2 emissions and income in the long-run according to time
series analysis These mixed results regarding the existence of EKC is expected as the economic development is not evenly distributed in the region
The long-run elasticity estimate of CO2emissions with respect
to energy consumption is positive at 1% significance level for the Philippines and Singapore; 5% significance level for Malaysia and Thailand and 10% significance level for Indonesia This finding in-dicates that higher energy consumption will result more carbon dioxide emissions and more polluted environment in these countries
The diagnostic tests results confirm the absence of serial cor-relation and heteroscedasticity in the estimated models The un-derlying models also pass diagnostic tests for normality and functional form The stability of short-run as well as long-run
co-efficients by testing the CUSUM and CUSUMSQ tests proposed by Brown et al.[48]were also tested.Fig 2presents that the stability
of coefficient estimates in all the cases is supported because the plots of both CUSUM and CUSUMSQ fall inside the critical bounds of 5% significance
Short-run estimation results in error-correction representation are provided inTable 5along with some diagnostic tests Although
a positive and negative coefficient for GDP and GDP2per capita are found in the cases of Malaysia, Singapore and Thailand, the results support the validity of EKC hypothesis in the short-run only in the case of Thailand as the related coefficients to LY and (LY)2are
sig-nificant at 1% and 5% significance levels respectively The short-run elasticity of real GDP per capita with respect to CO2emissions per capita is 3.2515e0.315 ln Y for Thailand The turning point of per capita real income is 10.32, compared to the highest value of ln Y for Thailand which is 7.866
Similar to the long-run results, in the cases of Indonesia and Philippines, short-run results indicated negative and positive
co-efficients for LY and (LY)2at 5% significance level respectively which
Table 3
The results of ARDL cointegration.
*, **, and *** Represent 10%, 5% and 1% level of significance, respectively.
a The critical values are obtained from Narayan [46, p 1988] , critical values for the bounds test: case III: unrestricted intercept and no trend.
Table 4
Long-run estimates based on selected ARDL models.
Diagnostic test statistics Serial correlationc2 (1) [p-value] Functional formc2 (1) [p-value] Normalityc2 (2) [p-value] Heteroscedasticity.c2 (1) [p-value]
Note: 1 *, **, and *** Represent 10%, 5% and 1% level of significance, respectively.
6 If the variable Y is measured in logs then exp(Y*) will yield the monetary value
B Saboori, J Sulaiman / Energy 55 (2013) 813e822 818
Trang 8-0.5 0.0 0.5 1.0 1.5
1976 1981 1986 1991 1996 2001 2006 2009
Plot of Cumulative Sum of Squares of Recursive Residuals Plot of Cumulative Sum of Recursive Residuals
-5
-10
-15
0
5
10
15
1976 1981 1986 1991 1996 2001 2006 2009
Thailand
Plot of Cumulative Sum of Squares of Recursive Residuals
-0.5 0.0 0.5 1.0 1.5
1975 1980 1985 1990 1995 2000 2005 2009
Plot of Cumulative Sum of Recursive Residuals
-5
-10
-15
-20
0
5
10
15
20
1975 1980 1985 1990 1995 2000 2005 2009
Singapore
Plot of Cumulative Sum of Squares of Recursive Residuals
-0.5 0.0 0.5 1.0 1.5
1973 1978 1983 1988 1993 1998 2003 2008 2009
Plot of Cumulative Sum of Recursive Residuals
-5
-10
-15
-20
0
5
10
15
20
1973 1978 1983 1988 1993 1998 2003 2008 2009
Philippines
Plot of Cumulative Sum of Squares of Recursive Residuals
-0.5 0.0 0.5 1.0 1.5
1974 1979 1984 1989 1994 1999 2004 2009
Plot of Cumulative Sum of Recursive Residuals
-5
-10
-15
0
5
10
15
1974 1979 1984 1989 1994 1999 2004 2009
Malaysia
Plot of Cumulative Sum of Squares of Recursive Residuals
-0.5 0.0 0.5 1.0 1.5
1974 1979 1984 1989 1994 1999 2004 2009
Plot of Cumulative Sum of Recursive Residuals
-5
-10
-15
-20
0
5
10
15
20
1974 1979 1984 1989 1994 1999 2004 2009
Fig 2 Plots of CUSUM and CUSUMSQ tests for the parameter stability The straight lines represent critical bounds at 5% significance level.
Trang 9resembles U-shaped relationship between carbon emissions and
economic growth Thisfinding is consistent with Nasir and Rehman
(2011), a study for Pakistan, which did not find an inverted
U-shaped relationship between economic growth and CO2emissions
in the short-run based on Johansen cointegration test This result
may be justified by the fact that EKC is a long-run phenomenon
[14]
There are significant coefficients at 1%, 5% and 10% significance
level for energy consumption with respect to CO2emissions in all
thefive selected counties This implies that energy consumption
plays a significant role in increasing CO2emissions in these
coun-tries in the short-run Comparing the long and short-run elasticities
of energy consumption variable with respect to carbon emissions
indicate that, the long-run elasticities are higher than the
short-run This implies that carbon emissions level, as a consequence of
energy consumption is found to increase over time in these ASEAN
countries
The existence of long-run relationship among carbon emissions,
economic growth and energy consumption suggests that there
must be Granger causality at least in one direction The results of
the causal relationship between the variables by using VECM based
Granger causality test are summarized inTable 6 The results show
that there are evidences of three long-run bi-directional Granger
causality relationships between the variables Thefirst one is a
bi-directional Granger causality between energy consumption and
CO2emissions in allfive ASEAN countries under study This finding
is in line with Apergis and Payne[31]; Pao and Tsai[33]; Al-Mulali
[19]and Pao et al.[53] This suggests that carbon emissions and
energy consumption are highly interrelated to each other
The second and third bi-directional Granger causality
relation-ships are between economic growth and CO2emissions and
eco-nomic growth and energy consumption in Indonesia, Malaysia and
Philippine This is similar to those arrived at by the previous
studies.7
There exists also evidence of short-run bi-directional Granger
causality between economic growth and CO2 emissions in
Indonesia, Singapore and Thailand Absent of short-run causality
from economic growth to CO2emissions in the cases of Malaysia
and Philippines implies that economic growth is not a proper
so-lution to reduce the levels of CO2emissions in the short-run in
these countries In the cases of Malaysia and Singapore there exists
Table 5
Short-run results based on selected ARDL models.
Indonesia: ECM ¼ LE þ 6.0508*LY 0.46624*(LY) 2 1.0395*LEN 12.9752
Malaysia: ECM ¼ LE 0.95232*LY þ 0.10012*(LY) 2 0.59484*LEN þ 1.8713
Philippines: ECM ¼ LE þ 206.6519*LY 14.5468*(LY) 2 5.2450*LEN 700.7801
Singapore: ECM ¼ LE 7.8326*LY þ 0.45256*(LY) 2 0.90304*LEN þ 38.3108
Thailand: ECM ¼ LE 9.0158*LY þ 0.60355*(LY) 2 1.1187*LEN þ 40.3575
Note: 1 ** and *** Represent 5% and 1% level of significance, respectively.
Table 6 Granger causality results.
Short-run Granger causality F-statistics [prob]
Long-run Granger causality
Indonesia
[0.0004]
0.734203 [0.4883]
0.358969 (2.650718)***
[0.0013]
[0.3686]
0.194581 (2.068479)**
(0.1405)
0.443666 [0.6458]
(2.417196)** Malaysia
[0.0975]
3.590767 [0.0400]
0.382612 (2.475320)**
[0.3043]
[0.0008]
0.718145 (3.430560)***
[0.0006]
2.773636 [0.0796]
(3.752216)*** Philippines
[0.0810]
5.285077 [0.1108]
0.216131 (1.788241)*
[0.0052]
[0.0002]
0.232788 (4.275188)***
[0.2585]
2.684992 [0.0683]
(1.942264)* Singapore
[0.0359]
5.811685 [0.0080]
0.342269 (2.072325)**
[0.0495]
[0.6674]
0.025102 (0.571669)
[0.0043]
0.291334 [0.7494]
(2.516405)** Thailand
[0.0005]
2.470372 [0.1016]
0.230009 (1.991027)*
[0.0002]
[0.0330]
0.052716 (0.700057)
[ 0.7161]
2.637201 [0.0881]
(1.867017)* Note: the null hypothesis is that there is no causal relationship between variables Values in brackets are p-values for Wald tests with F distribution ECt1represents the error-correction term lagged one period The optimal lag is based on AIC.D
first difference operator.
7
B Saboori, J Sulaiman / Energy 55 (2013) 813e822 820
Trang 10another short-run bi-directional Granger causality, between energy
consumption and CO2emissions This implies that an increase in
energy consumption may increase carbon emissions and vice versa
Energy consumption causes economic growth in the short-run in
Malaysia, Philippines and Thailand but the inverse is not true The
results of granger causality movements are summarized inFig 3
4 Conclusion
This study examines the cointegration and causal relationship
between economic growth, carbon dioxide (CO2) emissions and
energy consumption in five selected ASEAN countries namely
Indonesia, Malaysia, Philippines, Singapore and Thailand during
the period 1971e2008 The recently ARDL methodology
pro-posed by Pesaran and Shin [39] and Pesaran et al [40], and
Granger causality test based on VECM are employed The study
established cointegration relationship between carbon
emis-sions, energy consumption and economic growth in all the
countries with positive and statistically significant relationship
between carbon emissions and energy consumption in both short
and long-run
Production of industrial output and evolution toward export-oriented technologies in ASEAN countries put more pressures on the amount of energy consumed About 90% of the ASEANs primary commercial energy requirement is fulfilled by fossil fuels such as coal, oil, and gas This may lead to more emissions which in turn will make the need for pollution control actions more urgent Furthermore following Narayan and Narayan [56] the long-run elasticities of energy consumption variable with respect to carbon emissions are higher than the short-run elasticities This implies that carbon emissions level is found to increase in respect to energy consumption over time in the selected ASEAN countries
Under the long-run analysis, there is a positive long-run elasticity estimate of carbon emissions with respect to real GDP per capita and
a negative long-run elasticity estimate of carbon emissions with respect to the square of per capita real GDP at 1% significance level in Singapore, 10% significance level in the case of Thailand and statis-tically insignificance in the case of Malaysia The long-run results indicate that there is a negative long-run elasticity estimate of car-bon emissions with respect to real GDP per capita and a positive long-run elasticity estimate of carbon emissions with respect to the square of per capita real GDP at 10% significance level in the cases of Indonesia and Philippine Thefinding implies that over time income
Bi-directional long-run Granger causality Bi-directional short-run Granger causality Uni-directional long-run Granger causality Uni-directional short-run Granger causality
Y
Y
Philippines
Y
Singapore
Y
Thailand
Y
Fig 3 Granger causality relationship flows.