International Journal of Energy Economics and

Policy
ISSN: 2146-4553

available at http: www.econjournals.com

International Journal of Energy Economics and Policy, 2019, 9(5), 55-62.

Forecasting Carbon Dioxide Emission and Sustainable

Economy: Evidence and Policy Responses

Phatchapa Boontome1*, Apichit Therdyothin1, Jaruwan Chontanawat2

1
Division of Energy Management Technology, School of Energy, Environment and Materials, King Mongkut’s University of
Technology Thonburi, 126 Pracha Uthit Road., Bang Mod, Thung Khru, Bangkok,10140, Thailand, 2Division of Social Sciences
and Humanities, School of Liberal Arts, King Mongkut’s University of Technology Thonburi, 126 Pracha Uthit Road., Bang Mod,
Thung Khru, Bangkok,10140, Thailand. *Email: phatchapa.b@gmail.com

Received: 23 March 2019 Accepted: 28 June 2019 DOI: https://doi.org/10.32479/ijeep.7918

ABSTRACT
Forecasting CO2 emissions have been of importance as it could help the government to improve energy policies and plans. In this paper, we forecast
the future carbon dioxide emission (CO2) through estimating the short and long-run causal correlation between CO2 emission, economic growth (Y),
oil price (OP), consumption of renewable (RE), energy (E) in Thailand for the period 1990-2016 using autoregressive distributed lag approach. The
result indicates that in the long term, consumption of renewable, energy and oil price (OP) increase of 1% each decrease CO2 emission by 5.66%,
14.73% and 5.07% respectively. The result of forecasting CO2 emission base on variance decompositions found that in the future next 14-year decrease
CO2 emission 30.17%, which is higher than the target set to reduce CO2 emissions by 20-25% within 2030 year. The country should be adjust the
structure of energy use to reduce pollution.
Keywords: Forecasting, Carbon Dioxide Emission, Variance Decomposition
JEL Classifications: P28, Q42, Q43, Q47, Q48

1. INTRODUCTION global warming 20 years ago by signing and ratifying international

law related to this issue. Both the United Nations Convention on
Global economic fluctuations and energy prices, including global Climate Change (1994), the Kyoto Protocol (2002) and the Paris
warming and environmental impacts. All factors that affect and Agreement (2016). Thailand is a medium-level greenhouse gas
influence the situation of the energy use and supply of energy in emission of 350 million tons of carbon dioxide per year, accounting
the country. Scenario planning is one way of looking at the future. for 0.8% of the total, ranking 21st in the world. Thailand focuses
Analysis of factors that may affect future events, the linking of on energy security by distributing energy sources, reduce fossil
various factors, including the uncertainty that may occur, is able energy consumption and increasing the proportion of renewable
to cope with the changing world. Emerging developing countries energy consumption. Thailand has established a plan to reduce
play an important role in climate change because these countries greenhouse gases to at least 20-25% by 2030 (Banchanont, 2017).
generally have high economic growth affecting rapid pollution The government target low carbon along with the sustainability
emission. of economic growth. Therefore, the transition to low-carbon
technology may help to achieve reduces carbon dioxide emission
Thailand is one of the developing countries with the use of energy and sustainable economic growth (Zhao and Luo, 2018). In order
more steadily and affects carbon dioxide emissions increases to improving the government’s targets of CO2 emission. We will be
causing air pollution. Thailand has been aware of the problem of forecasting carbon dioxide emission for the next 13 years, whether

This Journal is licensed under a Creative Commons Attribution 4.0 International License

International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019 55

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

they can achieve the goals set in 2030 to improving energy policies Figure 1: Framework for time series analysis
and plans in the further.

2. LITERATURE REVIEW

The problem of greenhouse gas and energy consumption increases

have led to restore interest in forecasting of CO2 emissions in the
future. There are considerable amounts of literature that discussed
and reviewed this issue (Abdullah, 2015). According to Suarez
and Menendez (2015) study forecasting CO2 emissions, they plan
25% decrease in CO2 emissions by 2050 according to an economic
and population growth that is more consistent with recent global
trends. Wu et al. (2015) focus on forecasting CO2 emissions in the
BRICS countries found that economic growth has effect on the
CO2 emissions. Pao and Tsai, (2011) and Pao et al. (2012) study
modeling and forecasting the CO2 emissions, energy consumption,
and economic growth. They found of the inverted U-shaped
relationships of both emissions and energy consumption.

According to Lotfalipour et al. (2013) prediction of CO2 emissions

in Iran, they found that carbon dioxide emissions will increase
to 66% in 2020 compared to 2010. Appiah et al. (2019) report
that joint effect of energy intensity, economic progress and
industrialization at constant decrease emissions by 2.46% in
Uganda. According to He et al. (2019), Al Mamun et al. (2014),
Ahmed et al. (2017), Salahuddin et al. (2018), Sasana and Aminata
(2019) and Kalaycı and Hayaloglu, (2019) analyze the effect of
energy-based on economic growth to CO2 emissions. They found
that positive relationship between energy and economic impacted
CO2 emissions. According to Saudi et al. (2019), Appiah (2018), Economic growth (Y) measured in constant billion 2005 USD
Heidari et al. (2015), Lean and Smyth, (2010), Mirza and Kanwal $/Ktoe, renewable energy consumption (RE) measured in Ktoe,
(2017) and Alshehry and Belloumi (2015) they study relationship energy consumption (E) measured in Ktoe, Carbon dioxide
between energy consumption, renewable energy consumption, emissions (CO2) measured in MtCO2, and OP measured in USD
economic growth, and CO2 emissions. The results also showed $/Ktoe.
that renewable energy consumption impact to reduces carbon
dioxide emission but energy consumption and economic growth 3.2. Methodology
lead to increase CO2 emission. This paper aims to forecast carbon This study employs time series data to analyze this relationship,
dioxide emission and sustainable economy in Thailand evidence we following method framework proposed in Figure  1. Three
and policy responses. Therefore, we analyze that country will testing procedures were used in this study. After that, forecasting
achieve the goal of reducing greenhouse gases by at least 20-25% the CO2 emission base on Variance Decomposition and Impulse
by the year 2030 as the target or not. The results are to be used Response analysis.
in the policy planning of Thailand. Hence, this research will be Stage 1: Unit root test to determine the stationarity of the time
beneficial to national management and future applications. The series.
research process was as follows in Figure 1, framework for Time Stage 2: We carry out the test of cointegration by employing
Series Analysis. autoregressive distributed lag (ARDL) method developed
by Pesaran et  al. (2001) and followed by Johansen
3. DATA AND METHODOLOGY cointegration technique.
Stage 3: Causality test. If found that no causality can use estimate
We start by estimating the long-  and short-term relationship short term causality using standard Ordinary Least
between consumption of energy, renewable, oil price (OP), Squares (OLS)/Vector autoregression model (VAR), if
economic growth, and CO2 emissions. Then we forecasting CO2 found that causality can use estimate short-term and long-
emissions for the next 14 year. term causality error correction model (ECM).

3.1. Data 3.2.1 The description of the model

Annual data covering the period of 1990 to 2016 were obtained Stage 1: Stationary test
from World Development Indicators and Energy Policy and Firstly, the testing unit root examines the time-series variables
Planning Office. The empirical analysis includes five factors: for stationarity. We apply the Augmented Dickey and Fuller

56 International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

(ADF) and Phillips and Perron tests (PP) were performed to The ARDL approach developed by Pesaran et  al. (2001) was
test whether the data indicate a difference in stationarity or used to test cointegration. The ARDL bounds test is used to test
trend stationary, and to define the number of unit roots at their the long-term relationship among variables. This approach has
levels. Both tests are used to check the robustness of the results many econometric advantages compared to the standard Johansen
(Magazzino, 2015). We estimate the regression equation in the cointegration test. To demonstrate the ARDL modeling approach,
following manner: the following simple model can be considered:
n
∆yt =  +  t +  yt −1 + ∑ ∆y
i =1
i t −i + et  (1)
yt = α + βxt + δzt + et

The equation with β, δ and e represent short term dynamic of

Where α, β, µi, n and denote the intercept, the coefficient on the the model. The error correction version of the ARDL model is
time trend T, the coefficient on the lagged dependant variable, given by:
number of lags and random error, respectively. Phillips and
Perrom also suggested the method for unit root test and given the p p p

following equation: ∆yt =  0 + ∑

=i l =i l =i l
 i ∆yt − i + ∑  i ∆xt − i + ∑ ∆z
i t −i + 1 yt − l

∆yt = α + βt + θyt−1 + et (2) + 2 xt − l + 3 zt − l + ul

Stage 2: Testing for cointegration The equation with λ represents long term relationship. The null
The term cointegration basically refers to that one or more hypothesis in the equation is λ1 + λ2 + λ3 = 0, which means non-
linear combinations of time-series data are stationary even existence of long-term relationship (Ftiti et al., 2016).
though they are individually non-stationary (Chen et al., 2019).
Before proceeding with cointegration analysis, it was essential The ARDL approach leads to estimation of the following
to determine the optimal lag length using the Likelihood unrestricted ECM by the OLS Method (OLS) (Ftiti et al., 2016),
Ratio (LR), Schwartz Criterion (SC), and Akaike Information as presented by Equations (3) to (7):
Criterion (AIC) (Achour and Belloumi, 2016). The Johansen
cointegration test is designed to obtain likelihood-ratios. There
∑ ∑
p1 q1
∆yt=  0Y + iY ∆Yt − i + X jRE ∆REt − j
are two tests: the maximum eigenvalue test and the trace test. = i 1 = j 0

For both test statistics, the initial Johansen test is used for testing
∑ ∑
r1 s1
+  kE ∆Et − k + kCO 2 ∆CO2,t − l
the null hypothesis of no cointegration against the alternative of =k 0=l 0
cointegration. The tests differ in terms of alternative hypothesis
∑
g1
+  ∆OPt − m + Y Yt 1 +  RE REt 1
(Shahbaz et al., 2017), trace test, and maximum eigenvalue as m = 0 mOP 
follows: +E Et 1 + CO 2 CO2,t 1 +  OP OPt 1 + 1t (3)
n
trace =
−T ∑ (
ln 1 − ˆi2
i = r +1
) q2
∆REt =  0 RE + ∑
iRE ∆REt − j +
=j 1 =
p2
i 0 ∑ X jY ∆Yt − i

∑ ∑
r2 s2
Hypothesis as follows: H0: r ≤ k, H1 = r > k, k = 0,…, n +  kE ∆Et − k +
 ∆CO2,t − l
=k 0=l 0 klc

+∑
g2
λmax (r, r+1) = −Tln (1−λr+1)  ∆OPt − m + Y Yt 1 +  RE REt 1
m = 0 mOP 
Hypothesis as follows: H0: r = k, H1: r = k+1, k = 0,…, n +E Et 1 + CO 2 CO2,t 1 +  OP OPt 1 +  2t (4)

Where λi is the estimated ordered eigenvalue obtained from the

∑ ∑
r3 p3
∆Et =0E +  KE ∆Et − k + X jY ∆Yt − i
estimated matrix and T is the number of usable observations after = k 1= i 0
lag adjustment. The trace statistics test the null hypothesis that the
∑ ∑
r3 s3
+  iRE ∆REt − j +  ∆CO2,t − l
number of the distinct cointegrating vector (r) is less than or equal= k 0= l 0 lCO 2
to r against a general alternative. The maximal eigenvalue tests
∑
g3
+  ∆OPt − m + Y Yt 1 +  RE REt 1
the null hypothesis that the number of the cointegrating vector m = 0 mOP 
is r against the alternative of r + 1 (Esso and Keho, 2016). The +E Et 1 + CO 2 CO2,t 1 +  OP OPt 1 +  3t (5)
result of lag length criteria and Johansen test showed in Table 1
and Table 2, respectively.
∑ ∑
s4 p4
∆CO2,t =  0CO 2 + lCO 2 ∆CO2,t −1 + X ∆Yt − i
=l 1 = i 1 jY
Table 1: Result of lag length criteria
∑ ∑
q4 r4
Lag LR AIC SC +  iRE ∆REt − j + kE ∆Et − k
=j 0=k 0
1 169.19* –17.64* –16.17*
+∑
g4
(*) indicates lag order selected by the criterion, LR: Sequential modified Likelihood  mOP ∆OPt − m + Y Yt 1 +  RE REt 1
Ratio test statistic (each test at 5% level), AIC: Akaike information criterion, SC:
m=0 
Schwarz information criterion +E Et 1 + CO 2 CO2,t 1 +  OP OPt 1 +  4t (6)

International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019 57

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

A  significant t-statistic on the Error Correction Model (ECM)

∑ ∑
p5 q5
∆OPt =  0OP +
lOP ∆OPt − m + X jY ∆Yt − i indicates the presence of a long-term causality relationship, while
= m 1= i 0
the significant variables in first differences provides evidence of
∑ ∑
q5 s5
+  kE ∆Et − k +  ∆CO2,t −1 short-term causality relationships. The ECM reflects the speed
= k 0= l 0 lCO 2
of the adjustment and how quickly the variables return to the
+∑
g5
 mRE ∆REt − J + Y Yt 1 +  RE REt 1 long-term equilibrium with a statistically significant coefficient
m=0 
+E Et 1 + CO 2 CO2,t 1 +  OP OPt 1 +  5t (7) (Alshehry and Belloumi, 2015).

3.2.2. Forecasting CO2 emission and economic growth

Where Δ is the first difference and εjt, (j = 1, 2, 3, 4, 5) are white
After determining the directions of causality from the application
noise error terms. The existence of a cointegration relationship
of the VECM. We apply the generalised impulse responses and
between the variables from Equations (3) to (7) was investigated by
variance decomposition following Koop et al. (2015) and Pesaran
testing the significance of the lagged levels of variables using the
and Shin (Pesaran and Shin, 1998; Georgantopoulos, 2012)
computed F-statistic. Pesaran et al. (2001) and Brini et al. (2017)
innovative studies. Hence, the Impulse Response analysis to find
suggested testing H0: φY = γRE = ηE = λCO2 = θOP =0, which means
responsiveness outcome variable in the VAR when a shock is put to
that the absence of cointegration cannot be rejected, against the
error term. Consequently, unit shock is applied to each of variables
alternative, and H1: φY ≠ γRE ≠ ηE ≠ λCO2 ≠ θOP ≠ 0, which implies
in order to see its effects on VAR system (Appiah et al., 2019).
that the hypothesis of there being such a relationship cannot be
After that, this study predicts a causal link between economic
rejected. Following Pesaran et al. (2001) and Brini et al. (2017),
growth, renewable energy consumption, energy consumption,
the F-statistic used for this test has a non-standard asymptotic
CO2 emissions and OP in Thailand for the next 14-year period.
distribution and generates two sets of critical value bounds. The
lower critical value corresponds to the case where all variables are
I (0), and the upper critical value corresponds to the case where 4. EMPIRICAL RESULTS
all variables are I (1). If the computed F-statistic surpasses the
upper critical bound, then the null hypothesis of no cointegration 4.1. Results of Stationary Test
is rejected and it can be concluded that there is evidence of a We begin with the screening of influencing factors for model
long-term relationship. If it falls below the lower critical value, input. That regression results may be spurious if the estimated
the null hypothesis of no cointegration is not rejected, and if the variables are non-stationary and/or not cointegrated (Tang et al.,
F-statistic is between the lower and upper critical bounds, the 2016). Therefore, we testing for a unit root of each series is
result is inconclusive (Fatai et al., 2004). necessary. To investigate the order of integration, we began by
applying the ADF and PP. The ADF test and PP test show that all
Stage 3: Testing for causality variable causal factors are stationary at the First Difference I (1),
After the appearance of a long-term association amongst the as presented in Table 3.
variables, but it does not indicate the direction of causality. For this
aim, the Granger-causality tests based on vector error correction Table  3 showed that all factors are non-stationary at Level I
model (VECM) are carried out to the empirical analysis involved (0), it was found that those factors became stationary at first
testing both short-  and long-term causality between variables difference I (1). Once the factors were identified as stationary,
(Alshehry and Belloumi, 2015), applying the VECM approach they were taken for a co-integration test, as next step by Johansen
established in the following equation: Juselius and ARDL approach as shown reported in Tables 2 and
4, respectively.
 11, i 12, i 13, i 14, i 15, i 
  21, i  21, i  21, i  21, i  4.2. Results of the Cointegration Test
 21,i For estimate the long-term relationship through bound testing
 31,i 31, i 31, i 31, i 31, i 
 ∆Y   1    approach, there are essential tests which are used to select
 ∆RE      41,i  41, i  41, i  41, i  41, i  appropriate lag selection criteria. We employed LR, SC, and
   2   51,i
 51, i 51, i 51, i 51, i  AIC to identify appropriate lag length, as reported in Table  1.
∑
n
 ∆E  =  3  + The optimal lag length is found to be one, thus we performed the
  i =1
   ∆Yt − i    1   1 
 ∆CO2   4   ∆RE      Johansen cointegration test with this lag structure.
 ∆OP   5   t −i   2   2
 ∆Et − i  +  3   ECTt −1  +  3  We perform Johansen’s cointegration tests to assess the evidence
      of cointegration. In Table  2, both the results of trace tests and
 ∆CO2, t − i   4   4 
 ∆OPt − i   5   5  maximum eigenvalue tests unanimously point to the same
conclusion that there is at most one cointegrated relationship, at
 (8) the 1% level of significance. The results investigate each causal
factor was cointegrated at a confidence interval 95%.
Where ECTt−1 is the lagged error correction term derived from
the long-term cointegration functions; α1, α2, α3, α4, and α5 are In addition, we performed using the ARDL bounds test for
constant trends; γ1, γ2, γ3, γ4, and γ5 are the adjustment coefficients; compared to the standard Johansen cointegration test, we
and ε1, ε2, ε3, ε4, and ε5 are the serially uncorrelated error terms. investigate F-value to confirm the existence of cointegration

58 International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

Table 2: Results of Johansen’s cointegration test

Variables Hypothesized no. of cointegrating equations Trace statistic test Max‑eigen statistic MacKinnon critical value
1% 5%
∆ln (Y), None*** 85.35 34.67 39.37 33.87
∆ln (RE),
∆ln (E),
∆ln (CO2),
∆ln (OP)
At Most 1** 50.67 22.64 32.71 21.58
(***) denotes a significance, α=0.001, (**) denotes a significance, α=0.01, (*) denotes a significance, α=0.1

between variables for long term relationship. The results of bound Table 3: ADF and PP tests
F-statistics shown in Table 4. If the F-statistic is below the lower At 1st difference ADF Test PP tests
critical bound, I (0), then one accepts the null hypothesis of no Variables Value Value
cointegration between the variables which implies that the series ∆ln (Y) −3.56** −3.56*
are not cointegrated. However, if the F-statistic is below the ∆ln (RE) −4.94*** −5.36***
∆ln (E) −3.75* −3.85*
upper critical bound, I (1), then one rejects the null hypothesis ∆ln (CO2) −3.45* −3.48*
of no cointegration and therefore, concludes that the series are ∆ln (OP) −4.39* −4.38*
cointegrated. Furthermore, if the F-statistic falls in-between the Y is economic growth; RE is renewable energy consumption; E is energy consumption;
four critical bounds, the decision becomes inconclusive (Ang, CO2 is carbon dioxide emissions; OP is oil price. The values in this table are
2008). The calculated F statistic Fy (Y/OP, RE, E, CO2) = 2.79, Fop t‑statistics. (***) denotes a significance, α=0.001, (**) denotes a significance,
α=0.01, (*) denotes a significance, α=0.1, ∆ is the first difference, and ln is the natural
(OP/Y, RE, E, CO2) = 2.58, FE (E/OP, RE, Y, CO2) = 2.46, FCO2 logarithm. ADF: Augmented Dickey and Fuller, PP: Phillips and Perron
(CO2/OP, RE, E, Y) = 4.69, are greater than the upper bound critical
value of Pesaran et al. (2001) at the 10% significance level (3.52). Table 4: Results of bounds cointegration test – ARDL
This result indicates that the null hypothesis of cointegration was approach
rejected. FRE (RE/Y, OP, E, CO2) = 2.05 is the lower critical bound,
Equation F‑statistics
I (0), at the 10% significance level (2.45) then one accepts the FY (Y/OP, RE, E, CO2) 2.79*
null hypothesis of no cointegration between the variables which FOP (OP/Y, RE, E, CO2) 2.58*
implies that the series are not cointegrated. FRE (RE/Y, OP, E, CO2) 2.05
FE (E/OP, RE, Y, CO2) 2.46*
Both Johansen cointegration approach and the ARDL bounds test FCO2 (CO2/OP, RE, E, Y) 4.69*
Critical values of F‑statistics I (0)=2.45, I (1)=3.52
approach confirm that these variables form a long-term relationship
Critical values for K = 4 and n = 27 at 10%, (*) denotes a significance, α=0.1.
with each other; however, the presence of a long-term relationship
ARDL: Autoregressive distributed lag
does not imply causality. Therefore, we employed different variants
of the Granger causality test to study the direction of the causality
Table 5: Results of long‑and short‑term estimates of
between these variables (Heidari et al., 2015). Table 5 reported in the
ARDL model (Dependent Variable: CO2)
estimated long-term and short-term by the estimate ARDL model. In
Variable Coefficient P value
the long term, consumption of renewable, energy and OP increase
Long term estimates
of 1% each decrease CO2 emission by 5.66%, 14.73% and 5.07% ∆Y 1.1382 0.0000
respectively. In the short term, consumption of renewable, energy ∆RE −0.0566 0.3312
and OP increase of 1% each decrease CO2 emission by 10.08%, ∆E −0.1473 0.1365
26.22% and 9.03% respectively. But economic growth increase of ∆OP −0.0507 0.0420
1% each increase CO2 emission by 76.49%. Furthermore, the lagged ECTt−1 −1.7796 0.0000
Short term estimates
error correction term (ECTt−1) is of the right negative sign, which Constant −0.0133 0.0974
means that an equilibrium relationship could be restored any time Y 0.7649 0.0001
there are deviations. The coefficient of lag error correction term RE −0.1008 0.3329
(ECTt−1) −1.7796 implies that approximately 1% of the shocks to the E −0.2622 0.1339
system are restored in the next period. The same results suggest that OP −0.0903 0.0358
ARDL model (1, 1, 0, 0, 0); R2=0.83; Adj. R2=0.77; F‑Stats=3.58; Prob. (F‑Stats)=0.00;
a deviation from the long-term equilibrium level of CO2 emission DW=2.01; Normality test: Jarque‑Bera test=4.89; Hetero. Test=0.71;
in 1 year is corrected by 17.79% in the next year. ARDL: Autoregressive distributed lag

4.3. Results of the Granger Causality based on VECM calculate short term and long-term joint causality (Inglesi-Lotz,
approach 2016). Table 6 reported results from Granger Causality based on
After estimating the long-and short-term estimate, we find the VECM test, the results of the Wald-test are provided in Table 7
direction of causality by Granger causality base on the VECM and depicted in Figure 2 following below.
in order to find the direction of causality between variable. To
calculate short term causality, we apply Wald test to difference According to these results, the ECTt-1 coefficients of renewable
and lag difference coefficient of all independent variables. We also energy (RE), CO2 emissions (CO2) and OP are both comprised

International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019 59

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

Table 6: Result from Granger Causality based on VECM test

Dependent variable Short‑term causality Long‑term causality
∆Yt−1 ∆Et−1 ∆REt−1 ∆CO2, t−1 ∆OPt−1 ECMt−1
∆Yt ‑ –0.13 (0.7538) –0.03 (0.8516) 0.08 (0.8130) –0.10 (0.1672) –0.60 (0.2714)
∆Et 1.27 (0.0196)* ‑ –0.61 (0.0050)*** 0.63 (01214) –0.17 (0.0322)* –0.20 (0.5213)
∆REt 0.60 (0.2549) –0.24 (0.5564) ‑ –0.57 (0.2719) 0.22 (0.0161)* –0.35 (0.0768)*
∆CO2, t 0.76 (0.0408)* 0.51 (0.0970)* –0.48 (0.0118)* ‑ –0.12 (0.0850)* –1.53 (0.0072)***
∆OPt 0.98 (0.5258) 0.23 (0.8591) 0.50 (0.4630) –1.46 (0.2992) ‑ –0.48 (0.0796)*
Associated P value are provided in parentheses. (***) P<0.001; (**) P<0.05; (*) P<0.1. VECM: Vector error correction model

Table 7: Result from the wald test

dependent Variable Long‑term causality
∆Yt − 1 ECMt − 1 ∆Et − 1 ECMt − 1 ∆REt − 1 ECMt − 1 ∆CO2, i − 1 ∆OPt − 1 ECMt − 1
ECMt −1
∆Yt ‑ 0.64 (0.5366) 0.67 (0.5236) 0.68 (0.5190) 1.73 (0.2045)
∆Et 4.30 (0.0297)* ‑ −5.27 (0.0158)* 1.92 (0.1751) −4.24 (0.0309)*
∆REt 2.77 (0.0888)* 1.85 (0.1857) ‑ −2.66 (0.0971)* 4.74 (0.0222)*
∆CO2, t 8.07 (0.0031)*** 5.10 (0.0175)* −6.75 (0.0065)*** ‑ −8.11 (0.0031)***
∆OPt 1.72 (0.2056) 2.77 (0.0891)* 3.40 (0.0555)* −2.36 (0.1226) ‑
Associated P values are provided in parenthesis. *** P < 0.01; **P < 0.05; *P < 0.1

Figure 2: Causality graphs

between −1 and 0, which are statistically significant. Therefore, series of measures to prevent the further atrophy of environmental
there exists causality in the Granger sense running from Y to RE, pollution.
E, CO2 in the long-term. These results verify the cointegration test
analysis. In addition, there exists a long-term Granger causality In the short-term, there is a unidirectional causality running from
running from E, Y, and negative OP, RE to CO2. And then we RE to negative E, CO2 which means that use of renewable energy
found causality running from RE to OP and negative E, CO2. In effect in decreasing energy consumption from fossil and decreasing
the short-term, there is a uni-directional causality from running carbon emissions could promote the consumption of renewable
Y to E, CO2 at 5% level of significance. This may explain that energy. This result is consistent with the fact that Thailand is
the proportion of economic growth in energy consumption constantly increasing the use of renewable energy in recent years.
and CO2 emission are increasing, thus, the demand for energy In the long-term, renewable energy consumption causes negative
consumption also would increase. Simultaneous, the demand CO2 emissions. This result has been confirmed by the long-term
for renewable energy consumption also would increase too. In estimates coefficient in ARDL model showing that increasing
the short-term and long term, there is a unidirectional causality renewable energy consumption reduces of CO2 emissions. In the
running from E to CO2 at 1% level of significance which means short-term and long-term, there’s a unidirectional causality running
that increasing in energy consumption could promote the more from OP to RE, negative E, CO2 at 1% level of significance. This
energy efficiency and conservation policy. It can be explained that means that OP affects the use of energy, renewable energy and
with the deterioration of the environment, Thailand will take a CO2 in Thailand. The conclusion is consistent with the actual

60 International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

Table 8: Result from variance decompositions

Variance Decomposition of CO2
Period Standard error Y RE E CO2 OP
3 0.074458 5.774378 0.065181 5.777781 62.63071 25.75195
7 0.094857 5.062426 0.267859 25.98938 46.63098 22.04935
14 0.139598 3.629858 7.959773 29.58542 30.17206 28.65289
Economic growth (Y), renewable energy consumption (RE), energy consumption (E), Carbon dioxide emissions (CO2) and OP

situation in Thailand that the OP expensive effect to decreasing use bi-directional causality relationship between renewable energy
of energy and pollution emissions but when OPs fluctuate effect consumption and CO2 emission, OP respectively. As well as we
to the production of more renewable energy. Figure 2 sums up found that strong bi-directional causality relationship between
short-term and long-term Granger causalities between variables. OP and energy consumption, CO2 emission respectively. This
indicates each consumption of renewable, energy, CO2 emission,
4.4. Result of forecasting CO2 emission and economic OP is highly interrelated to each other.
growth
Table 8 shows the estimations of the variance decomposition for We have predicted the CO2 emission based on VAR model within
CO2 emission. These results indicate that form of CO2 emission 2017-2030 for next 14 year. The results show CO2 emission is
initially explains relatively of the future variation in CO2 emission. forecasted to decrease 30.17% in the next year 14. Thailand
In the short run, that is year 3 about 62.63% of CO2 emission are has implemented an action plan to reduce greenhouse gases by
caused by its own standard innovation shock. CO2 emission reacts 20-25 percent by 2030, aiming to reduce the use of fossil energy
by 5.77%, 0.06% and 25.75% when a one standard deviation and use more renewable energy to be environmentally friendly.
change is imputed in economic growth, renewable energy Increase the proportion of renewable energy use and sustainable
consumption, energy consumption and OP, respectively. development to achieve the goal of reducing greenhouse gas and
sustainable development in year 2030. The results found that in
In the long term, that is year 14, impulse or innovation or shock to the next 14 years, carbon dioxide emissions were 30.17%. Which
CO2 emission account for 30.17 percent variation of the fluctuation has more carbon dioxide emissions than the target set to reduce
in CO2 emission, shock to economic growth, renewable energy carbon dioxide emissions by 20-25% in 2030.
consumption, energy consumption, OP can cause 3.62%, 7.95%,
29.58% and 28.65% respectively, fluctuation in CO2 emission. As We should encourage more renewable energy production to
a result, total fluctuation become 100 percent. However, as the reduce carbon emissions. Therefore, improving energy efficiency
forecast horizon widens the explanatory of CO2 emission decreases and reduce pollution the long-term environmental would support
by 30.17% at year 14. The results indicate that a decrease in enhancing the sustainability of further economic growth and
economic growth in Thailand would lead to a decrease in carbon target in Thailand. Additionally, this study’s it attempts to pave a
dioxide emissions as 1% decrease in economic growth would give guideline for future research and in different research contexts.
abate to 5.06% in carbon dioxide in year 7. Results of the impulse Moreover, it will help national policy planning in the future.
response function the negative effects of the various independent However, should be the difference various concepts are to produce
variables on the environment of Thailand. The result shows that a useful result to achieve goals sustainable policy.
in the next 14  years, Thailand cannot reduce carbon dioxide
emissions to meet the goals of the policy that in 2030 will reduce REFERENCES
greenhouse gas emissions by 20-25%.
Abdullah, L., Pauzi, H.M. (2015), Methods in forecasting carbon dioxide
5. CONCLUSION AND POLICY emissions: A decade review. Jurnal Tehnologi, 75, 67-82.
Achour, H., Belloumi, M. (2016), Investigating the causal relationship
RECOMMENDATIONS between transport infrastructure, transport energy consumption and
economic growth in Tunisia. Renewable and Sustainable Energy
The main objective of this study is to forecast the CO2 emission in Reviews, 56, 988-998.
the future (for the next 14 years) that are sufficient to the goals set Ahmed, K., Rehman, M.U., Ozturk, I. (2017), What drives carbon
by the government or not, using Johansen cointegration and ARDL dioxide emissions in the long-run? Evidence from selected South
models for test of long-term cointegration. Granger causality test Asian countries. Renewable and Sustainable Energy Reviews, 70,
for direction causality and apply to forecast based on VAR model. 1142-1153.
We analyzed the causal relationship between CO2 emissions, Al Mamun, M., Sohag, K., Mia, M.A.H., Uddin, G.S., Ozturk, I. (2014),
Regional differences in the dynamic linkage between CO2 emissions,
renewable energy consumption, energy consumption, OPs, and
sectoral output and economic growth. Renewable and Sustainable
economic growth in Thailand from 1990 to 2016. The results of
Energy Reviews, 38, 1-11.
estimate cointegrating relationships between variable, there exist Alshehry, A.S., Belloumi, M. (2015), Energy consumption, carbon
positive and statistically significant coefficients of CO2 emission dioxide emissions and economic growth: The case of Saudi Arabia.
with respect to economic growth, renewable energy consumption, Renewable and Sustainable Energy Reviews, 41, 237-247.
energy consumption, OP is short and long-term. The Granger Ang, J.B. (2008), Economic development, pollutant emissions and
causality testing results show that there exists long-term and strong energy consumption in Malaysia. Journal of Policy Modeling,

International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019 61

 Boontome, et al.:Forecasting Carbon Dioxide Emission and Sustainable Economy: Evidence and Policy Responses

30(2), 271-278. and output in ASEAN. Applied Energy, 87(6), 1858-1864.

Appiah, K., Du, J., Yeboah, M., Appiah, R. (2019), Causal relationship Lotfalipour, M.R., Falahi, M.A., Bastam, M. (2013), Prediction of CO2
between industrialization, energy intensity, economic growth emissions in Iran using Grey and ARIMA models. International
and carbon dioxide emissions: Recent evidence from Uganda. Journal of Energy Economics and Policy, 3(3), 229-237.
International Journal of Energy Economics and Policy, 9(2), 237-245. Magazzino, C. (2015), Economic growth, CO2 emissions and energy
Appiah, M.O. (2018), Investigating the multivariate Granger causality use in Israel. International Journal of Sustainable Development and
between energy consumption, economic growth and CO2 emissions World Ecology, 22(1), 89-97.
in Ghana. Energy Policy, 112, 198-208. Mirza, F.M., Kanwal, A. (2017), Energy consumption, carbon emissions
Banchanont, P. (2017), 5 Things to Know about Thailand and the Paris and economic growth in Pakistan: Dynamic causality analysis.
Agreement and the “Global Warming” Problem. Available from: Renewable and Sustainable Energy Reviews, 72, 1233-1240.
https://www.bbc.com/thai/international-40137393. Pao, H.T., Fu, H.C., Tseng, C.L. (2012), Forecasting of CO2 emissions,
Brini, R., Amara, M., Jemmali, H. (2017), Renewable energy energy consumption and economic growth in China using an
consumption, international trade, oil price and economic growth improved grey model. Energy, 40(1), 400-409.
inter-linkages: The case of Tunisia. Renewable and Sustainable Pao, H.T., Tsai, C.M. (2011), Modeling and forecasting the CO2 emissions,
Energy Reviews, 76, 620-627. energy consumption, and economic growth in Brazil. Energy, 36(5),
Chen, Y., Wang, Z., Zhong, Z. (2019), CO2 emissions, economic growth, 2450-2458.
renewable and non-renewable energy production and foreign trade Pesaran, H.H., Shin, Y. (1998), Generalized impulse response analysis in
in China. Renewable Energy, 131, 208-216. linear multivariate models. Economics Letters, 58(1), 17-29.
Esso, L.J., Keho, Y. (2016), Energy consumption, economic growth and Pesaran, M., Shin, Y., Smith, R. (2001), Bounds testing approaches to
carbon emissions: Cointegration and causality evidence from selected the analysis of level relationships. Journal of Applied Econometrics,
African countries. Energy, 114, 492-497. 16(3), 289-326.
Fatai, K., Oxley, L., Scrimgeour, F.G. (2004), Modelling the causal Salahuddin, M., Alam, K., Ozturk, I., Sohag, K. (2018), The effects of
relationship between energy consumption and GDP in New electricity consumption, economic growth, financial development
Zealand, Australia, India, Indonesia, the Philippines and Thailand. and foreign direct investment on CO2 emissions in Kuwait.
Mathematics and Computers in Simulation, 64(3), 431-445. Renewable and Sustainable Energy Reviews, 81, 2002-2010.
Ftiti, Z., Guesmi, K., Teulon, F., Chouachi, S. (2016), Relationship Sasana, H., Aminata, J. (2019), Energy subsidy, energy consumption,
between crude oil prices and economic growth In selected OPEC economic growth, and carbon dioxide emission: Indonesian case
Countries. The Journal of Applied Business Research, 32(1), 11-22. studies. International Journal of Energy Economics and Policy,
Georgantopoulos, A.G. (2012), Electricity consumption and economic 9(2), 117-122.
growth: Analysis and forecasts using VAR/VEC approach for greece Saudi, M.H.M., Sinaga, O., Jabarullah, N.H. (2019), The role of
with capital formation. International Journal of Energy Economics renewable, non-renewable energy consumption and technology
and Policy, 2(4), 263-278. innovation in testing environmental Kuznets Curve in Malaysia.
He, F.S., Gan, G.G.G., Al-Mulali, U., Adebola, S.S. (2019), The influences International Journal of Energy Economics and Policy, 9(1), 299-307.
of economic indicators on environmental pollution in Malaysia. Shahbaz, M., Sarwar, S., Chen, W., Malik, M.N. (2017), Dynamics of
International Journal of Energy Economics and Policy, 9(2), 123-131. electricity consumption, oil price and economic growth: Global
Heidari, H., Katircioglu, S.T., Saeidpour, L. (2015), Economic growth, perspective. Energy Policy, 108, 256-270.
CO 2 emissions, and energy consumption in the five ASEAN Suarez, R.P., Menendez, A.J.L. (2015), Growing green? Forecasting CO2
countries. International Journal of Electrical Power and Energy emissions with environmental Kuznets Curves and logistic growth
Systems, 64, 785-791. models. Environmental Science and Policy, 54, 428-437.
Inglesi-Lotz, R. (2016), The impact of renewable energy consumption Tang, C.F., Tan, B.W., Ozturk, I. (2016), Energy consumption and
to economic growth: A panel data application. Energy Economics, economic growth in Vietnam. Renewable and Sustainable Energy
53, 58-63. Reviews, 54, 1506-1514.
Kalaycı, C., Hayaloglu, P. (2019), The impact of economic globalization Wu, L., Liu, S., Liu, D., Fang, Z., Xu, H. (2015), Modelling and
on CO2 emissions: The case of NAFTA countries. International forecasting CO2 emissions in the BRICS (Brazil, Russia, India,
Journal of Energy Economics and Policy, 9(1), 356-360. China, and South Africa) countries using a novel multi-variable grey
Koop, G., Pesaran, M.H., Potter, S.M. (1996), Impulse response analysis model. Energy, 79, 489-495.
in nonlinear multivariate models. Journal of Econometrics, 74(1), Zhao, X., Luo, D. (2018), Forecasting fossil energy consumption structure
119-147. toward low-carbon and sustainable economy in China: Evidence and
Lean, H.H., Smyth, R. (2010), CO2 emissions, electricity consumption policy responses. Energy Strategy Reviews, 22, 303-312.

62 International Journal of Energy Economics and Policy | Vol 9 • Issue 5 • 2019