You are on page 1of 10

2nd Reading

September 25, 2014 14:44 WSPC/2335-6804 1450014

International Journal of Energy and Statistics


Vol. 2, No. 3 (2014) 197–206
c Institute for International Energy Studies
DOI: 10.1142/S2335680414500148

CO2 emissions and output in Iran:


Environmental Kuznets Curve approaches
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

Afshin Javan∗ and Alireza Ghanbari†


∗Modelling
and Forecasting Analyst,
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

OPEC, Vienna, Austria


ajavan@opec.org
†Financial Expert,

National Iranian Gas Export Company, Iran


alirezaghanbari@yahoo.com

Received 20 June 2014


Revised 26 August 2014
Accepted 27 August 2014
Published 30 September 2014

This paper evaluates the dynamic causal relationships between CO2 emissions and
energy consumption in Iran for the period 1965–2008 based on the EKC Hypothesis.
The VECM technique is exploited for the empirical analysis. This study has been com-
pleted in two stages. In the first stage, we examine how the variables are related in
the long-run, and at the second stage, survey the dynamic causal relationships between
the variables. The empirical results have provided evidence for the existence of a strong
long-run relationship between CO2 emissions, energy consumption and real GDP in
Iran. Moreover, the results indicate that the increase of GDP and energy consumption
simultaneously will increases CO2 emissions and also based on preliminary results, the
energy consumption effects are greater than GDP effects on CO2 emission. A short
run Granger causality test suggests bidirectional causality between CO2 emissions and
energy consumption, unidirectional causality from GDP to CO2 emissions, and uni-
directional causality from GDP to energy consumption. In addition, these causalities
confirmed that in the long run there are unidirectional relations from CO2 emissions to
GDP and from energy consumption to GDP.

Keywords: CO2 emissions; Environmental Kuznets Curve hypothesis; Vector Error-


Correction model; Iran.

Nomenclature
CO2 : Carbon dioxide.
EKC : Environmental Kuznets Curve.
VECM : Vector Error-Correction model.
ECM : Error Correction model.
GDP : Gross Domestic Product.

197
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

198 A. Javan & A. Ghanbari

GHG : Greenhouse Gas.


Gt : Real GDP.
Ct : CO2 emissions.
Et : Commercial energy consumption.
G2 : Square of Real GDP.
ADF : Augmented Dickey–Fuller.
ECT : Error Correction Term.
PP : Phillips–Perron.
AIC : Akaike’s information criterion.
SBC : Schwarz criterion.
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

HQ : Hannan–Quinn criterion.
εt : Error term.
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

αi (i = 1, . . . , 3) : Model Parameters.
β11 , β21 , β31 , β41 : VECM parameters.
I(1) : Unit Root.

1. Introduction
Carbon dioxide is the main GHGs generated by an upward trend during the last
decade in Iran and total CO2 emission from energy consumption has been reported
as 624.86 million tons in the year 2011. The use of fossil fuels as the main source
of energy supply in Iran has transformed Iran into one of the 20 countries which
have a contribution of 75% of total GHG generation [1, 2]. Figure 1 presents the 20
major countries in the list of CO2 emissions for 2009. There is a relation between
CO2 emissions and energy consumption, and since GDP is dependent on energy
consumption, there is a correlation between CO2 emissions, GDP and energy con-
sumption as presented in Fig. 2. This simple graphical analysis reveals that GDP
and energy consumption do not have a linear relationship and that a quadratic spec-
ification is required based on the EKC Hypothesis. Also it appears that the levels
of real GDP, energy consumption and CO2 emissions have significantly increased
over time.
Recently there has been increased attention on energy related environmental
issues (see for example [1, 3]). Iran’s rank in terms of CO2 emissions in the world
was at 27th place in 1965 and it deteriorated up to 11th place in 2008. The Iranian
share of the total world CO2 emissions was 1.61% in 2008. This means that CO2
emissions in Iran are growing faster than ever before. In this paper, we examine
the dynamic relationships between GDP, CO2 emissions and energy consumption
in Iran.
The remainder of this paper is organized as follows. A brief literature review is
presented in Sec. 2 which is followed by an introduction to the data, definitions and
a discussion of time series properties of the variables in Sec. 3. Section 4 presents the
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

CO2 emissions and energy consumption in Iran 199


Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

Fig. 1. CO2 emissions — Iran vs. World.

300,000 5000
4500
250,000 4000
200,000 3500
3000
150,000 2500
2000
100,000 1500
50,000 1000
500
0 0
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009

Co2 Energy Consumpon GDP (Per Capita)

Fig. 2. Iran variables trend.

empirical results whilst policy implications are considered in Sec. 5 which concludes
the paper.

2. Literature Review
Until recently there have been two parallel literatures on the relationship between
GDP, energy consumption and GHG emissions. The first part of studies on the
relationship between GDP and energy consumption has focused on the economic
activity energy consumption nexus since the seminal study by Kraft and Kraft [4].
A voluminous Granger causality literature has emerged examining the link between
economic growth and energy consumption (see [5] for a recent review). One con-
strain of this literature is that many of the studies consider the relationship between
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

200 A. Javan & A. Ghanbari

economic growth and energy consumption in a functional framework and thus suffer
from omitted variables bias [6, 7].
The second part of the research is related to pollution and GDP nexus that call
on the EKC hypothesis. This hypothesis states that as income increases, emissions
increase as well until some threshold level of income is reached after which emissions
begin to decline. Dinda [7] and Stern [8] review and critique against EKC literature.
The EKC hypothesis specifies emissions as a function of income, which presumes
unidirectional causality runs from GDP to CO2 emissions. However, it is conceiv-
able that causation could run from emissions to income whereby emissions occur
in the production process and, as a consequence, income increases. Considering
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

this point, few studies have examined the direction of Granger causality between
economic growth and environmental pollution [9–12]. While these studies are a def-
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

inite improvement on the standard EKC literature in the sense that they model the
time series dynamics, they continue to suffer from the problem of omitted variables
bias [13]. A marriage of these two literatures is considered whereby the relation-
ship between GDP, energy consumption and CO2 emissions are evaluated. There
are studies for developed countries, such as France [14] and the United States [15],
developing countries, such as China [16], Malaysia [17] and Turkey [18, 19], as well
as the oil-rich OPEC countries [20].
The results from these studies however are mixed. For example, [15, 19] found
an unidirectional long run Granger causality relationship from energy consump-
tion to pollution emissions, while [18] found a bidirectional long run and short
run Granger causality relationship between economic growth and pollution emis-
sions. Zhang and Cheng [16] found unidirectional Granger causality running from
economic growth to energy consumption and energy consumption to pollution
emissions in the long run, while [14] found unidirectional Granger causality run-
ning from economic growth to energy consumption and pollution emissions in the
long-run. There exist a more limited number of studies which consider the link
between economic growth and environmental degradation in Iran. Fetros et al.
[21] test the relationship between economic growth and pollution in Iran. They
use the Toda-Yamamoto approach for determining the causality direction amongst
variables. The results indicate that there are three unidirectional relations, from
CO2 emissions to per capita GDP, from CO2 emissions to per capita energy con-
sumption, and from per capita energy consumption to water pollution. Sharzei
[22] investigates the effect of energy consumption and GDP on carbon emissions
in Iran during the period 1978–2005. He investigates whether there is a Granger
causality relationship between income, energy consumption, and carbon emission
by including labor and gross fixed capital formation in the model. Behbodi and
Collogues [23] studied relations between energy consumption, economic growth
and CO2 emissions based on time series covering the period 1967–2004 by apply-
ing an ECM model. They concluded that there is a significant positive relation
between energy consumption, economic growth, foreign trade, population and CO2
emissions.
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

CO2 emissions and energy consumption in Iran 201

3. Model, Data and Methodology


3.1. Model and data
Based on the basic theoretical arguments, the relationship between GDP, CO2
emissions, and energy consumption are as follows:

Gt = f (Ct , Et ), (1)

where f is functional form. In contrast, based on the EKC hypothesis, GDP and
CO2 emissions have a nonlinear quadratic relationship. Hence, the long run rela-
tionship between CO2 emissions, energy consumption and GDP is as follows:
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

Ct = α0 + α1 Gt − α2 G2t + α3 Et + εt , (2)
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

where αi (i = 1, . . . , 3) are the model’s parameters. In Eq. (2), based on EKC


hypothesis is it expected that CO2 emissions have a positive relationship with
real GDP and a negative relationship with squared real GDP. It is also expected
that the positive sign indicates a direct relation between energy consumption and
CO2 emissions (if we consume more hence produce more CO2 ). Respectively the
GDP coefficients have long-run elasticities with CO2 emissions and also squared
real GDP and energy consumption. Therefore the expectation in the model was
that all variables would be co-integrated and all coefficients would be significant.
However, if coefficients of Eq. (2) are statistically insignificant, it is concluded that
with increase in GDP and CO2 emissions, the curve between CO2 emissions and
GDP increases monotonically.
The variables are CO2 emissions (measured in million tons), commercial energy
consumption (measured in million tons) and GDP (measured in million $ at base
year 2000). In line with the parametric assumptions, all series are transformed into
natural logarithms. Thus, the series can be interpreted in growth terms after taking
the first difference. Annual time series covering the period 1965–2008 are used in
this paper [26].

3.2. Methodology
The testing procedure in this paper involves three main steps as follows:

(i) Integration test via two unit root tests — ADF test and PP test.
(ii) Test for co-integration using the Johansen’s [24] approach.
(iii) Test for causality by employing the appropriate types of causality tests.

According to Engle and Granger [25], co-integrated variables must have an error
correction representation in which an ECT must be incorporated into the model.
Accordingly, a VECM is formulated to reintroduce the information lost in the dif-
ferencing process, thereby allowing for long-run equilibrium as well as short-run
dynamics. For the 4-variable case with one co-integrated relationship, the VECM
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

202 A. Javan & A. Ghanbari

can be expressed as follows:


p−1
 p−1

∆Ct = α1 + α11 ECT t−1 + φ1j ∆Ct−j + θ1j ∆Et−j
j=1 j=1

p−1
 p−1

+ Ψ1j ∆Gt−j + δ1j ∆G2t−j + ε1t , (3)
j=1 j=1

p−1
 p−1

∆Et = α2 + α21 ECT t−1 + φ2j ∆Ct−j + θ2j ∆Et−j
j=1 j=1
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

p−1
 p−1

δ2j ∆G2t−j + ε2t ,
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

+ Ψ2j ∆Gt−j + (4)


j=1 j=1

p−1
 p−1

∆Gt = α3 + α31 ECT t−1 + φ3j ∆Ct−j + θ3j ∆Et−j
j=1 j=1

p−1
 p−1

+ Ψ3j ∆Gt−j + δ3j ∆G2t−j + ε3t , (5)
j=1 j=1

p−1
 p−1

∆G2t = α4 + α41 ECT t−1 + φ4j ∆Ct−j + θ4j ∆Et−j
j=1 j=1

p−1
 p−1

+ Ψ4j ∆Gt−j + δ4j ∆G2t−j + ε4t , (6)
j=1 j=1

where ECT t−1 = Ct−1 + (β21 /β11 )Et−1 + (β31 /β11 )Gt−1 + (β41 /β11 )G2t−1 is the
normalized co-integrated equation and θij = βi+1,j /β1j , ψij = βi+2,j /β1j and δij =
βi+3,j /β1j . There are two sources of causation, i.e., through the ECT if α = 0 or
through the lagged dynamic terms. The ECT measures the long-run equilibrium
relationship while the coefficients on lagged difference terms indicate the short-
run dynamics. The statistical significance of the coefficients associated with ECT
provides evidence of an error correction mechanism that drives the variables back
to their long-run relationship.
Given the two different sources of causality, we can perform two different causal-
ity tests, i.e., short-run Granger non-causality test and long-run weak erogeneity
test. In Eq. (3), to test ∆Gt and ∆G2t do not Granger cause ∆Ct in the short
run, we examine the significance of the lagged dynamic terms by testing the null
H0 : all Ψ1j = all δ1j = 0 using the Wald test. Rejection of the null implies GDP
growth Granger causes pollution growth in the short run. To test the null that
pollution growth does not Granger cause GDP growth, a different procedure is
necessary since there are two variables which indicate GDP growth in the system,
i.e., ∆Gt and ∆G2t . Using Eqs. (5) and (6), we formulate an unrestricted VECM
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

CO2 emissions and energy consumption in Iran 203

by imposing the restrictions all φ3j = 0 and all φ4j = 0. Similar to the con-
ventional F-test, a likelihood ratio test can be performed on the determinants of
residual variance of these two models. Rejection of the null implies pollution growth
Granger causes GDP growth in the short run. The weak erogeneity test, which is a
notion of long-run non-causality test, requires satisfying the null H0 : α11 = 0 for
non-causality from long-run equilibrium deviation to ∆Ct . For non-causality from
long-run equilibrium deviation to ∆Gt and ∆G2t , this requires non-rejection of the
null H0 : α31 = α41 = 0. The hypothesis testing is based on a likelihood ratio test
which follows a χ2 distribution [24].
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

4. Empirical Results
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

According to the results of unit root tests in the Table 1, based on the ADF and PP
tests all model variables are non-stationary in their levels but become stationary
after taking the first difference. We conclude that all series are I(1) and significant
at the 5% level.
After finding the integrating order, the two-step co-integration procedure has
been employed. In the first stage, AIC, SBC and HQ information criteria are utilized
to select the optimum lag length of the model. The results are presented in Table 2.
Since the objective is to select the optimal order, it is important that at this stage
we select the optimal order will not exceed it. The lags 0, 1, 2, 3 have been estimated
over the sample. However, AIC and HQ criteria implied that the order is 2 and SBC
is 1. As a result, we choose first order of the model as the optimal lag.
Given that the variables have common integration properties we proceed to
test for the presence of a common trend, or equivalently, a long-run co-integrating
relationship between them. Base on eigenvalue and maximum eigenvalue statistics
we conclude that there is just one long run relation among variables (Table 3).

Table 1. Unit root tests.


ADF PP
Level Lst difference Level Lst difference
Et −3.31 −4.74∗ −2.62 −4.58∗
Ct −3.43 −4.80∗ −2.68 −4.65∗
Gt −3.11 −3.70∗ −2.02 −3.21∗
∗ Indicates 5% level of significance.

Table 2. Selecting the order of the model.

Order AIC SBI HQ


0 −7.34 −7.17 −7.28
1 −17.90- −17.06∗ −17.60
2 −18.31∗ −16.81 −17.76∗
3 −18.23 −16.05 −17.43
∗ Indicates significance at 1% level.
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

204 A. Javan & A. Ghanbari

Table 3. Cointegration test results (Johansen).

Eigenvalue Maximum eigenvalue statistics


Hypothesized of
cointegrating equations Statistic Critical value Statistic Critical value
R= 0 30.47 27.42 56.28 48.88
R= 1 18.82 21.12 25.82 31.54
R= 2 6.72 14.88 6.96 17.86
R= 3 0.25 8.07 0.25 8.07

Table 4. Co-integrating vector.


Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

Intercept G E G2
Coefficient 0.75 −0.23 1.01 −0.04
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

t-statistic −2.17 −97.42 3.23

Table 5. Causality tests.

Short run Granger Long-run weak


Hypothesis non causality test exogeneity test
H0 : ∆C→E 5.16∗ 0.184
H0 : ∆E→C 5.52∗ 0.154
H0 : ∆C→G,G2 2.99 33.44∗
H0 : ∆G,G2 →C 7.19∗ 0.154
H0 : ∆E→G,G2 3.42 33.44∗
H0 : ∆G,G2 →E 6.86∗ 0.184
∗ Indicated 5% level of significance.

All coefficients are statistically significant at the 5% level and (in the long-run
relationship) have the expected signs in Table 4 as presented the co-integrating
equation. Ct is positively related to Et but negatively related to G2t . The results
provide some support for our argument to examine these variables under an inte-
grated framework.
Co-integration implies the existence of causality, at least in one direction. How-
ever, it does not indicate the direction of the causal relationship between variables.
Also we perform the ECM-based causality tests to shed light on the direction of
causality. The results prepared in Table 5 show a bidirectional causality running
between CO2 emissions and energy consumption in the short run. Also there is uni-
directional causality running from GDP growth to CO2 emissions and from GDP
to energy consumption in the short run. The results also suggest CO2 emissions
causal GDP and GDP causal energy consumption in long run.

5. Conclusion
Iran is a developing economy which is one of the major pollutant countries in the
world. As mentioned above Iran has a strong effect on global warming. In this paper,
we examine the dynamic relationship between CO2 emissions, energy consumption,
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

CO2 emissions and energy consumption in Iran 205

and GDP in Iran over the period 1965–2008 using a multivariate VECM. The
empirical results for the case of Iran suggest the existence of a robust long-run
relationship between the variables. Based on these empirical results the elasticity
of CO2 emissions with respect to GDP is inelastic; however the elasticity of CO2
emissions with respect to energy consumption is elastic in the long-run. The results
indicate that more energy consumption results in more CO2 emissions, and CO2
emissions and GDP have a quadratic relationship in the long run.
To complement the findings of the co-integration analysis, we also perform
causality tests to shed light on the causal links of GDP, energy consumption and
CO2 emissions. The results indicate that GDP causes to increase both energy con-
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

sumption and CO2 emissions in the short run. The results also display a bidirec-
tional causality running between CO2 emissions and energy consumption in the
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

short run and point to a unidirectional causality running from energy consumption
to GDP growth in the long run. According to the empirical results, we believe that
Iran should design new environmental policies to reduce environmental degradation
and apply new methods for decreasing rational energy consumption and the direct
relation between CO2 emissions and GDP.

References
[1] Bagheri Moghaddam, N., Mousavi, S. M., Nasiri, M., Moallemi, E. A. and Yousefdehi,
H. (2011). Wind energy status of Iran: Evaluating Iran’s technological capability
in manufacturing windturbines. Renewable and Sustainable Energy Reviews, 15(8),
4200–4211.
[2] Parker, L. and Blodgett, J. (2005). Greenhouse gas emission: Conflicting situations,
conflicting perspectives. Congressional Research Service, Library of Congress.
[3] Silva, E. S. (2013). A combination forecast for energy-related CO2 emissions in the
United States. International Journal of Energy and Statistics, 1(4), 269–279.
[4] Kraft, J. and Kraft, A. (1978). Note and Comments: On the relationship between
energy and GNP. The Journal of Energy and Development, 3(2), 401–403.
[5] Apergis, N. and Payne, J. E. (2009). CO2 emissions, energy usage and output in
Central America. Energy Economics, 37(8), 3282–3286.
[6] Stern, D. I. (1993). Energy and Economic Growth in the USA: A Multivariate
Approach. Energy Economics, 15(2), 137–150.
[7] Dinda, S. (2004). Environmental Kuznets Curve Hypothesis: A Survey. Ecological
Economics, 49(4), 431–455
[8] Stern, D. I. (2000). A Multivariate cointegration analysis of the role of energy in the
US economy. Energy Economics, 22(2), 267–283.
[9] Coondoo, D. and Dinda, S. (2002). Causality between income and emission: A country
group-specific econometric analysis. Ecological Economics, 40(3), 351–367.
[10] Dinda, S. and Coondoo, D. (2006). Income and emission: A panel data-based coin-
tegration analysis. Ecological Economics, 57(2), 167–181.
[11] Akbostanci, E., Türüt-AsIk, S. and Tunç, G. I. (2009). The relationship between
income and environment in Turkey: Is there an environmental Kuznets curve? Energy
Policy, 37(3), 861–867.
[12] Lee, C. C. and Lee, J. D. (2009). Income and CO2 emissions: Evidence from panel
unit root and cointegration tests. Energy Policy, 37(2), 413–423.
2nd Reading
September 25, 2014 14:44 WSPC/2335-6804 1450014

206 A. Javan & A. Ghanbari

[13] Lean, H. H. and Smyth, R. (2010). CO2 emissions, electricity consumption and output
in ASEAN. Applied Energy, 87(6), 1858–1864.
[14] Ang, J. B. (2007). CO2 emissions, energy consumption, and output in France. Energy
Policy, 35(10), 4772–4778.
[15] Soytas, U., Sari, R. and Ewing, B. T. (2007). Energy consumption, income, and
carbon emissions in the United States. Ecological Economics, 62(3–4), 482–489.
[16] Zang, H. (1998). The stability of the Kuznets Curve: Some further evidence. Appli-
cation of Economics Letters, 5(3), 131–133.
[17] Ang, J. (2008). Private investment and financial sector policies in developing coun-
tries. Monash Economics Working Papers 07/08, Monash University, Department of
Economics.
[18] Halicioglu, F. (2009). An econometric study of CO2 emissions, energy consumption,
Int. J. Energy Stat. 2014.02:197-206. Downloaded from www.worldscientific.com

income and foreign trade in Turkey. Energy Policy, 37(3), 1156–1164.


[19] Sari, R. and Soytas, U. (2009). Are global warming and economic growth compatible?
by MCMASTER UNIVERSITY on 01/15/15. For personal use only.

Evidence from five OPEC countries? Applied Energy, 86(10), 1887–1893.


[20] Förster, A., Schöner, R., Förster, H. J., Norden, B., Blaschke, A. W., Luckert, J.,
Beutler, G., Gaupp, R. and Rhede, D. (2010). Reservoir characterization of a CO2
storage aquifer: The Upper Triassic Stuttgart Formation in the Northeast German
Basin. Marine and Petroleum Geology, 27(10), 2156–2172.
[21] Sharzehi, G. and Molaei, M. (2008). Derivation of shadow price for CO2 gas emission
using distance function approach. Iranian Economic Review, 13(21), 123–135.
[22] Behboudi, D., Mohammadzadeh, P. and Jebraeili, S. (2010). An assessment of the
relationship between energy consumption and GDP in developing and developed
countries. Quarterly Energy Economics Review, 6(23), 1–21.
[23] Johansen, S. (1988). Statistical analysis of cointegration vectors. Journal of Economic
Dynamics and Control, 12(2–3), 231–254.
[24] Ang, J. B. (2007). CO2 emissions, energy consumption, and output in France. Energy
Policy, 35(10), 4772–4778.
[25] Engle, R. F. and Granger, C. W. J. (1987). Co-integration and error correction:
Representation, estimation, and testing. Econometrica, 55(2), 251–276.

You might also like