Accepted Manuscript

Modelling the CO2 emissions and Economic growth in Croatia: Is there any
Environmental Kuznets Curve?

Najid Ahmad, Liangsheng Du, Jiye Lu, Jianlin Wang, Hong-Zhou Li, Muhammad
Zaffar Hashmi

PII: S0360-5442(16)31911-9

DOI: 10.1016/j.energy.2016.12.106

Reference: EGY 10108

To appear in: Energy

Received Date: 19 April 2016

Revised Date: 01 December 2016

Accepted Date: 26 December 2016

Please cite this article as: Najid Ahmad, Liangsheng Du, Jiye Lu, Jianlin Wang, Hong-Zhou Li,
Muhammad Zaffar Hashmi, Modelling the CO2 emissions and Economic growth in Croatia: Is there
any Environmental Kuznets Curve?, Energy (2016), doi: 10.1016/j.energy.2016.12.106

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Highlights
 Environmental Kuznets Curve hypothesis was testing using ARDL approach for Croatia.

 Results confirm the validity of EKC for Croatia in long run.

 There was bi-directional causality between CO2 emissions and income in short run.

 There was uni-directional causality from income to CO2 emissions in long run.
 ACCEPTED MANUSCRIPT

1*Najid Ahmad, 2*Liangsheng Du, 3Jiye Lu, 4Jianlin Wang, 5Hong-Zhou Li, 6Muhammad Zaffar
Hashmi
1,2,3Schoolof Economics, Dongbei University of Finance and Economics, Dalian, Liaoning
Province, 116025, China
4,5Center
for Industrial and Business Organization, Dongbei University of Finance and
Economics, Dalian, Liaoning Province, 116025, China
6Department of Meteorology, COMSATS Institute of Information Technology, Islamabad,
Pakistan
Corresponding authors’ email addresses:
1*najid_2iqbal@yahoo.com and 2*duliangsheng@vip.sina.com
 ACCEPTED MANUSCRIPT

Modelling the CO2 emissions and Economic growth in Croatia: Is there any

Environmental Kuznets Curve?

Abstract
This paper investigates the existence of environmental Kuznets Curve in Croatia for the period

of 1992Q1-2011Q1. To fulfil to goals of the paper, Autoregressive Distributed Lag (ARDL)

and VECM method has been applied. Results show the existence of inverted U-shape relation

between CO2 emissions and economic growth in long run that is the validity of EKC. Granger

causality based on VECM approach shows bi-directional causality between CO2 emissions and

economic growth in short run and uni-directional causality from economic growth to CO2

emissions in long run. DOLS and FMOLS results confirm the robustness of long run results.

Variance decomposition and Impulse response also show similar results. Beauty of the paper

is the consistency of results from different techniques.

Keywords: ARDL approach, Croatia, Kuznets Curve, Causality

 ACCEPTED MANUSCRIPT

1. Introduction

Global warming is one of the biggest challenge that can destroy human race from the face of

earth. Environmentalists and environmental economists argued that CO2 is the main source of

this warming [1,2] and high carbon emissions is because of high growth [3,60].

Intergovernmental panel on climate change (IPCC) has pointed out that the climatic change is

raising sea level, melting ice from glaciers and causing extreme weather events that can destroy

50 percent coastal areas’ population [4,5]. Stern [6] projection has shown that the economic

loss from global warming will be higher than the World War I and World War II. IPCC has

also reported that greenhouse gas is increasing rapidly from last two decades and CO2

emissions is responsible for greenhouse effects [7]. This greenhouse gas emissions has

increased the average temperature of the earth [71] and it (greenhouse gas emissions) will

continue growing by 1.7 times till 2030 [8]. Increasing threats of climate change and global

warming has attracted researchers’ attention to focus on before it leads to serious destruction.

It has been viewed that energy consumption is main source of CO2 emissions [62]. So,

suggestion can be given to reduce energy consumption to overcome CO2 emissions problem

[9]. However, energy reduction will put negative pressure on the economic growth of an

economy that is obviously not good for the country.

Three decades back, environmental economists started arguing that environmental problems’

solution is in economic growth [11]. Following Kuznets1, Grossman and Krueger [13, 68, 69]

pointed out that similar Kuznets curve can be found between income and environmental

1 In 1955, Nobel Prize winner, Kuznets, pointed that at initial level, income and income
inequality has positive relation but after the specific level (turning point) this relation turns to
inverse. Thus, income and income inequality has inverted U-shape relation. This notation is
called as Kuznets Curve [12].
 ACCEPTED MANUSCRIPT

degradation and they called it Environmental Kuznets Curve (EKC). Idea of inverted U-Shape

or Environmental Kuznets Curve states that initially, increase in economic growth will lead to

environmental pressure but after the specific level, rise in economic growth will reduce

environmental pressure. In other words, EKC states that there is an inverted U-shape relation

between environmental degradation and economic growth. At initial stage, economic growth

will raise CO2 emissions but after the specific level (turning point), this relation will turn to

inverse. So, after the specific period of time, increase in economic growth will decrease CO2

emissions. In this way, economic growth itself is the solution for clean environment. Knowing

such a nice solution in the form of economic growth, there is growing literature to test EKC

since last two decades [4, 14-33, 59, 67, 70 and others]2. However, validity of EKC remain

debated issue for developing and developed countries as EKC results are very sensitive to

country, variables and time span [17].

If we talk about Croatia, It is an emerging economy and member of European Union (EU) since

2013. Being the part of EU, Croatia has become the part of package plan of 2020 target for

climate and energy. EU has set targets to reduce CO2 emissions like 20% CO2 emissions’

reduction till 2020 and 60-80% till 2050 [45]. For this, Governments are implementing tax or

fee on CO2 emissions. The money collected from the fee/tax will be spent on carbon dioxide

reduction plans. However, introduction and implementing of CO2 tax on vehicles does not

influence the emissions in Croatia as people’s purchasing power is not high enough to replace

old cars with new one [45]. Further, according to Energy Strategy of Croatia, electricity

consumption will rise from 2% to 18% [44] and it can cause CO2 emissions in the environment.

In the world development indicators, it has been shown that Croatia can face environmental

challenges because of its rapid growth and CO2 emissions. Knowing the importance of topic

2 For detail, see literature review section and appendix.

 ACCEPTED MANUSCRIPT

and need for Croatia, there was a need to test EKC to offer policy suggestion to overcome

environmental problems. Unfortunately, there is no study available where EKC has been tested

for Croatia. So, task of the present study is to fill the gap by testing EKC for Croatia. Up to our

knowledge, it is the first attempt to test Environmental Kuznets Curve for Croatia.

This study is parallel to recent studies [20, 24, and 46] for model construction to test EKC for

Croatia. Autoregressive Distributed lag (ARDL) by Pesaran et al. [34] has been used because

of its advantages3. Different diagnostic tests like CUSUM and CUSUMSQ for coefficients

stability, residuals’ distribution, functional form of the model, heteroskadesticity test (ARCH),

serial correlation are used to confirm model perfection. Dynamic ordinary least square (DOLS)

and fully modified ordinary least square (FMOLS) are used for robustness check. These

estimators (FMOLS and DOLS) are free from serial correlation, small sample bias and

endogenity issue [37-38]. ARDL offers co-integration among variables and does not explain

causality in short and long run. So, vector error correction model (VECM) has been applied for

causality. VECM shows causality within the sample and does not explain causality out of

sample [40]. So, variance decomposition and impulse response function (IRF) has been utilized

to check causal relation out of sample. Rest of the paper is structured as follows: section 2 for

literature review, section 3 is for data, method and results and section 4 concludes the paper.

2. Literature Review

There is growing literature to test environmental Kuznets curve since last two decades,

therefore, it is not easy to review all work. However, extensive review for the latest and relevant

work is presented in below lines to show previous work and their conclusion from results. For

example, Pao and Tsai [22] use annual data of GDP, CO2 emission for Brazil to test EKC.

ARDL approach confirmed that GDP was having positive coefficient while GDP square was

3 ARDL advantages and its detail is given in Methodology and Results section.
 ACCEPTED MANUSCRIPT

with negative coefficient in long run that was the confirmation of EKC in Brazil. For Malaysia,

Saboori and Sulaiman [26] use time series data of energy consumption, CO2 emissions and

economic growth to test the validity of EKC. They fail to confirm the validity of EKC when

energy consumption was used at aggregate level. However, they found the evidence of EKC at

disaggregated energy. Further, they were unable to find any evidence of EKC in short run from

different techniques and marked EKC as long run phenomena. They also found bi-directional

causality between economic growth and CO2 emissions. For Russia, Pao et al. [23] examine

CO2 emissions, economic growth and energy consumption relation for the period of 1990-

2007. They did not find the evidence of EKC for Russia and suggested energy conservation to

overcome environmental pollution. For Pakistan, Nasir and Rehman [19] explore the relation

between energy, CO2 emissions, economic growth and foreign trade. They confirmed the

existence of EKC in the long run with Johansen cointegration.

If we talk about China, Wang et al. [31] investigated the relation between economic growth,

CO2 emissions and energy consumption by utilizing 28 Chinese province data. They found U-

shape relation between economic growth and CO2 emissions that confirmed that EKC is not

valid in China. Bidirectional causality was also found between energy consumption, economic

growth and CO2 emissions. There is another recent study by Hao et al., [63] who tested the

existence of Environmental Kuznets Curve for China for coal consumption. 29 province data

has been used for the period of 1995-2012. They took cubic form of income in the model and

confirmed the validity of EKC for China. For Malaysia, Saboori et al. [24] explore the relation

between CO2 emissions and economic growth. ARDL approach confirmed the validity of

inverted U-shape curve in short and long run. Granger causality tests show absence of causality

in short run while there was causality from growth to CO2 emissions in long run. On the other

hand, Ozturk and Acaravci [20] use ARDL approach to test the relation between economic

growth, CO2 emissions, energy consumption and employment for Turkey. CO2 emissions was
 ACCEPTED MANUSCRIPT

having negative while energy consumption effect was positive in explanation of economic

growth. Causality results show absence of causality. Further, they fail to find the validity of

EKC and conclude that energy conservation and CO2 reduction policy will not have an adverse

effect on the economic growth of Turkey. Contrary, Ozturk and Acaravci [21] confirm the

validity of EKC for Turkey.

Apergis and Payne [16] examined the relation between CO2 emissions, economic growth and

energy consumption for 11 commonwealth countries and they confirmed the validity of EKC.

They concluded by saying that economic growth can serve to fight with environmental issues.

On the other hand, Jaunky [17] tested environmental Kuznets curve for 36 high income

countries by applying panel co-integration and GMM. At individual country’s analysis, he

found the validity of EKC for few countries and in panel case, there was no signal for EKC.

Similarly, Acaravci and Ozturk [14] investigated the relation between CO2 emissions,

economic growth and energy consumption for nineteen European countries and found that

EKC was not valid in most of the countries. Saboori and Sulaiman [25] explore the relation

between CO2 emissions, economic growth and energy consumption for five ASEAN countries

i.e. Malaysia, Thailand, Indonesia, Philippines and Singapore. With the help of ARDL, they

found the support of EKC for Thailand and Singapore in long run and this relation was weak

(insignificant) in Malaysia. Short run results only confirmed EKC for Thailand. Zaman et al.

[59] investigated EKC for three diversified region of the world namely high income OECD

and non-OECD countries, East Asia and Pacific and, European Union for the period of 2005-

2013. Being member of European Union, Croatia was included in the study. They confirm the

validity of Environmental Kuznets Curve in the panel of these regions. On the other hand,

Halicioglu and Ketenci [61] explore the relation among CO2 emissions, energy, economic

growth and trade openness for 15 transitions countries. ARDL and GMM confirmed the

validity of EKC for only three countries.

 ACCEPTED MANUSCRIPT

For Romania, Shahbaz et al. [30] explore the dynamic relation between CO2 emissions, energy

consumption and economic growth. ARDL confirm the validity of EKC in short and long run.

For India, Tiwari et al. [42] investigate the relation between CO2 emissions, coal consumption,

GDP growth and trade openness. They confirm the validity of EKC for short and long run with

ARDL approach. For Turkey, Yavuz [32] tested the validity of EKC by using CO2 emissions,

economic growth and energy consumption data for the period of 1960 to 2007. He confirmed

EKC validity in long run. However, Akbostanci et al. [15] found positive monotonic relation

between CO2 emissions and economic growth for Turkey and conclude that economic growth

is not sufficient to control environmental pollution. Above literature is witness to show that

different researchers find different results and thus, conclusion. Main reasons behind

conflicting results can be variables choice, time span and country’s economic situation.

3. Data, Method and Results

3.1 data and model

In order to test Environmental Kuznets Curve for Croatia, quarterly data has been used for the

period of 1992Q1 to 2011Q1 according to the data availability. CO2 emissions is in metric tons

while GDP is in billion US$ (constant 2005) and are retrieved from World Development

Indicators (WDI) and Economic Intelligence Unit (EIU) respectively. Variables are in

logarithms for smoothness purpose4. General form of the model can be written as:

CO2  F (GDP, GDPSQ, M) (1)

Here CO2 is carbon dioxide emissions, GDP is gross domestic product, GDPSQ is GDP Square.

M are other explanatory variables that can influence CO2 emissions. Knowing the objectives

4In what follows, analysis were conducted in natural log as well as non-log forms and their
results were qualitatively similar. We present the results for natural log form by following
reviewer 3 suggestion “to reduce data’s variations”, however, results for non-log form are
available on request.
 ACCEPTED MANUSCRIPT

of study, we will focus on testing cointegration and causality between economic growth and

CO2 emissions by considering reduced form of the model preferable as it allows to measure

the direct and indirect relation between the variables of concern. Thus, inclusion of the

additional variables may deviate from the primary goal of the paper. Further, reduced form of

the model eliminate the risk of loss of important variables from the analysis as most of the time

data availability become issue in the construction of possible determinants. Reduced form of

the data also reduce the risk of analytical freedom5. Thus, Environmental Kuznets Curve

general form can be as:

CO 2t   0  1 (GDP)t   2 (GDPSQ)t   t (2)

CO2 is carbon dioxide emissions, GDP is gross domestic product, GDPSQ is GDP Square, t is

time period, B0 is constant, B1 is coefficient of GDP, B2 is the coefficient of GDP square,  is

white noise error term. Here, if B1= B2=0, it will show level relation. If B1<0, B2=0 or B1>0,

B2=0, there will be monotonically decreasing and increasing linear relationship respectively.

What about if B1<0, B2>0, in this situation, there will be U-shape relation and if there is B1>0,

B2<0, there will be inverted U-shape relation that is EKC. Turning point of real income will be

as GDP*= – B1/2B2. As the value of GDP is in logs. So, exp(GDP*) will give the monetary

value representing the Peak of EKC.

3.2. Methodology and Results

3.2.1. ARDL for Cointegration and short, and long run estimates

Autoregressive Distributed Lag (ARDL) model by Pesaran et al. [34] has been used to test

cointegration. In this approach, Pesaran et al. [34] solved the non-stationary issue in time series

5 See List and Gallet [70] and Saboori et al. [24] for more detail about the reduced form of the

model and its importance.

 ACCEPTED MANUSCRIPT

data and make this model irrespective to stationary properties of the data. ARDL is applicable

if all variables are stationary at level i.e. I(0) or purely I(1) or mixture of both I(0) and I(1).

Generally, it is rare that any variable (s) move to second difference i.e. I(2) and particularly, in

case of small sample, it is hard to decide if any variable is second difference stationary.

However, if any variable turns to second difference, solution lies to take it’s difference and

make it first difference stationary. Then, ARDL can be used for I(0), I(1) or mixture of I(0) and

I(1) variables. Thus, ARDL offers package solution for all kind of variables and help to avoid

traditional unit root testing6. Whereas, OLS is only suitable if all variables are stationary at

level. Engle and Granger [35] technique is for two variables and order of integration should be

same i.e I(1). Johansen Co-integration [36] has its own limitations as it can be used only for

large sample size and variables should be integrated of same order i.e. I(1).On the other hand,

For ARDL, Monto Carlo Simulation has confirmed that ARDL technique is superior and

provide unbiased results for small sample size [34, 50, 51]. Further, in ARDL, model can be

easily estimated if all explanatory variables are endogenous [34]. Endogenity is the biggest

problem and the best way to eradicate endogenity can be the introduction of lags by making

model dynamic [52]. ARDL approach overcome this issue by making the model dynamic [34].

Owing these advantages, we use ARDL approach to investigate the co-integration between

CO2 emissions and economic growth for Croatia.

6We confirmed that none of our variable was second difference stationary to cover Ouattara
[58] comment that ARDL is not applicable for I(2) as critical bounds by Pesaran et al. [34]
are for I(0) and I(1) variables. To conserve space, we did not report unit roots detail,
however, are available on request.
 ACCEPTED MANUSCRIPT

Autoregressive Distributed Lag representation for long run will be as:

n n n
CO2t   3   1k CO2t  k    2 k GDPt  k   3k GDPSQt  k 1CO2 CO2t 1   2CO2 GDPt 1   3CO2 GDPSQt 1   3t (3)
k 1 k 0 k 0
n n n
GDPt   4   1k GDPt  k    2 k CO2t  k    3k GDPSQt  k  1GDP GDPt 1   2GDP CO2t 1   3GDP GDPSQt 1   4t (4)
k 1 k 0 k 0
n n n
GDPSQt   5   1k GDPSQt  k    2 k GDPt  k    3k CO2t  1GDPSQ GDPSQt 1   2GDPSQ GDPt 1   3GDPSQ CO2t 1   5t (5)
k 1 k 0 k 0

Here in equation 3, α3 is constant, εt is white noise error term, β1 β2 β3 are the error correction

dynamics while other coefficients ∆1CO2, ∆2CO2, ∆3CO2 are for long run. Same can be interpreted

for equation 4 and equation 5. For ARDL, first task is to estimate equation (let say equation 3)

by ordinary least square. In equation 3, null hypothesis of no co-integration is H0:

∆1CO2=∆2CO2=∆3CO2=0 against alternative H1:∆1CO2≠∆2CO2≠∆3CO2≠0. Wald test (F-statistics)

is used to test the existence of co-integration among variables. Critical values for this F-

statistics are given by Pesaran et al. [34]. They introduce two types of bounds. One is lower

bounds that consider all variables are as I(0) and other is upper bounds that takes all variables

as I(1). If calculated F-Statistics is higher than the upper bounds, it means, we can reject null

of no cointegration and concludes the existence of cointegration. If F-statistics is below the

lower bounds, it means that there is no long run relation among the variables. If F-statistics is

between lower and upper bounds, then results are inconclusive. In this situation, Banerjee et

al. [53] stated that negative and significance error correction will confirm the long run relation

among variables [53-54]. Knowing, this F-statistics is very sensitive to order of the variables

and number of lags [64], we utilize VAR to find maximum lag order that was 2 through AIC

and SBC as they are ideal for small sample [34, 65]. We move to check F-statistics that was

responsible to show cointegration among the variables. Table 1 shows that F-statistics (3.18)

is between lower and upper bounds showing inconclusive results. In this situation, error

correction term was very useful tool to check the existence of long run relationship. If error

correction term is negative and significance, it will confirm the long run relationship among
 ACCEPTED MANUSCRIPT

variables [53-54]. Table 2 shows that error correction term is negative (-0.053) and highly

significance that confirm the validity of long run relation for equation 3. Once, long run relation

has been confirmed, next step was to estimate error correction model (ECM) for equation 3, 4

and 5. Error correction general form will be as equation 6, 7 and 8.

n n n
CO2t   3   1k CO2t  k    2 k GDPt  k    3k GDPSQt  k  6 ECTt 1   3t (6)
k 1 k 1 k 1
n n n
GDPt   4   1k GDPt  k    2 k CO2t  k    3k GDPSQt  k   7 ECTt 1   4t (7)
k 1 k 1 k 1
n n n
GDPSQt   5   1k GDPSQt  k    2 k GDPt  k    3k CO2t  8 ECTt 1   5t (8)
k 1 k 1 k 1

Here ECT is error correction term that shows speed of adjustment. It shows model converge to

long run equilibrium path in case of disturbance in short run.  6,  7 and  8 are coefficients

of error correct term in equation 6, 7, 8 respectively. They are responsible to show adjustment

speed in which system will converge to long run equilibrium path. Long run and short run

coefficients are reported in table 2. Long run results show that GDP has positive coefficient

(6.465) and it is significance at 5% level of significance while GDP square coefficient is

negative (-0.832) and is significant at 5% significance level. So, our long run results show the

situation as B1>0, B2<0, that is the confirmation for the existence of Environmental Kuznets

Curve for Croatia. We plotted the graph (Fig. 1) from long run equation. The graph clearly

shows that initially, rise in economic growth is raising CO2 emissions but after the period of

time (long run), this relation is inverse i.e. further increase in economic growth is reducing CO2

emissions and graph is like inverted U-shape. Our results are line with Saboori, Sulaiman, &

Mohd [24] for Malaysia, Wang et al. [31] for China, Ozturk and Acaravci [21] for Turkey,

Apergis and Payne [16] for 11 commonwealth countries of independent states and Yavuz [32]

for Turkey. However, these results differ from Pao et al. [23] for Russia, Acaravci and Ozturk

[14] for nineteen European countries, Ozturk and Acaravci [20] for Turkey. Conflicting results
 ACCEPTED MANUSCRIPT

reason can be sensitivity of EKC to country, variables choice, time span, sample size and

methods.

Short run results show that GDP is having negative sign and GDP square is with positive sign

and, both coefficients are statistically insignificant. It reconfirms that EKC is long run

phenomena rather than short run and is valid in Croatia as the growth and development is long

run process. Thus, economy takes time to reach to specific level where further increase in

economic growth will decrease pollution that is possible in long run. These results are in line

with Dinda [66] and Nasir and Rehman [19] who stated that EKC is long run phenomena and

its validity should be judged in long run. We find the speed of adjustment (ECT) is having

negative sign and it is highly significant. ECT coefficient (-0.053) confirms that if there would

be disturbance in the short run equilibrium, model will converge to long run equilibrium path

with 5.3% adjustment speed on quarterly basis. Diagnostic results in table 3 confirm that there

is no serial correlation, model is correctly specified, residuals are normally distributed, and

model is free from heteroskadesticity problem. CUSM (Fig.2) and CUSMSQ (Fig.3) shows

that variables in the model are stable over the long period of time as critical bounds are within

the bounds. Thus, overall, model is perfect and reliable for policy purpose.

3.2.2. DOLS and FMOLS for robustness

Although ARDL results confirm the validity of EKC in long run and different diagnostic tests

showed model perfection, however, we use dynamic ordinary least square and fully modified

ordinary least square to check the robustness of our long run results to make our work more

interesting and novel. DOLS and FMOLS estimators are free from serial correlation,

endogenity issue and provide unbiased results in small sample [37, 39]. Table 4, results from

DOLS and FMOLS show GDP elasticity ranges between 5.11-5.13 and statistically

significance at 1% level of significance. GDP square coefficient in both cases is -0.63 and is
 ACCEPTED MANUSCRIPT

statistically significance at 1% level of significance. Both DOLS and FMOLS reconfirm the

validity of EKC in long run as in both cases GDP>0, and GDPSQ<0. Thus, our long run results

are robust.

3.2.3. The VECM Granger causality analysis

ARDL approach was ideal to test co-integration among the variables and thus, to test the

validity of EKC for Croatia. However, it does not explain causality between variables i.e. which

variable is causing which variable. Granger [55] approach can help to show which variable is

responsible to cause which variable in short and long run. In the absence of cointegration,

conventional VAR can be used to check causality between variables. However, in the presence

of cointegration, VAR can produce misleading results [25]. For cointegrated series, it is better

to take the difference of the variables to make them stationary and then use for VECM [56]. If

there is the conformity of the cointegration among variables, we obtain residuals from long run

equation and include the lagged residuals as additional independent variable. Then, we can

employ VECM approach. The negative and significant ECT coefficient shows long run causal

relation while short run causality can be checked via F-statistics of Wald test.

General form of the granger causality test would be as:

CO2t  b1  C11,i C12,i C13,i  CO2t i  1  1t 

    q       
(1  L) GDPt   b2    (1  L) C21,i C22,i C23,i  GDPt i    2    ECTt 1    2t  (9)
GDP 2  b  i 1 C C C  GDP 2     3t 
 t   3  31,i 32,i 33,i  t i   3

Where (1-L) is lag operator, ECT is lagged error correction term.  t ’s are uncorrelated random

disturbance with zero mean and C’s are parameters to be estimated. Lag length has been chosen

via AIC and SBC criterion.

 ACCEPTED MANUSCRIPT

Granger causality results show several interesting facts as: When CO2 emissions is dependent

variables, GDP coefficient is positive while GDP square coefficient is negative and they are

statistically significant at 1% level of significance in short run. The negative and highly

significance ECT confirms long run relation among variables. It shows the disturbance in the

system will be adjusted 68% in first quarter that is quite fast. In other words, disturbance in the

system will take less than a year to reach its equilibrium path in long run. Further, when GDP

was dependent variable, CO2 emissions and GDP square coefficients were positive and

significant at 1% level of significance in short run. In third equation, CO2 emissions was having

negative relation with GDP square while GDP was with positive impact on its square in short

run. In each cases (GDP and its square as dependent variable separately), error correction term

was having positive sign and statistically significant. It confirms the absence of long run

cointegration. In other words, it (ECT) will push the system away from equilibrium path in

case of disturbance in short run. To sum up, we find bi-directional causality between CO2

emissions and economic growth in short run. It means CO2 emissions and economic growth

are interdependent on each other. There was also uni-directional causality from economic

growth to CO2 emissions in long run.

3.2.4. Variance decomposition and Impulse response function

Vector error correction model offers granger causality within the sample and does not explain

out of sample causality [40]. In other words, it cannot direct us to check the behaviour of one

variable in the response of innovations of other variable in future. Variable response to other

variable beyond the sample can be seen through variance decomposition and impulse response

function [40]. So, variance decomposition and impulse response function has been used to

check out of sample causality. Engle and Granger [35] stated that variance decomposition

provide better results in VAR environment. Thus, VAR has been utilized for this purpose.

Variance decomposition explains how much one variable contribute in the explanation of other
 ACCEPTED MANUSCRIPT

variable while impulse response function explains how much one variable response to other

variable shocks including its own shock [56].

Variance decomposition results (table 5) show that in short run (period 1-5), CO2 emissions

explains itself around 93.5% and GDP contribution in the explanation of carbon dioxide is

5.9%. GDP square contribution is insignificant. However, from short run to long run (5-10

periods), it can be seen that CO2 emissions explains itself around 78.7%. GDP explains CO2

emissions 20.9% and GDPSQ contribution is around 0.60%. When we move further, we can

see GDPSQ contribution is increasing significantly and GDP contribution in the explanation

of CO2 emissions is not increasing. Variance decomposition of GDP shows that at initial

periods (1-5), GDP explains itself around 81.9% while CO2 emissions explains GDP around

17.8% and GDP square contribution is insignificant. It can be seen on period 20 that GDP

explains itself around 70% and CO2 emissions contribute 28.9% in the explanation of GDP.

Similarly, variance decomposition of GDP square shows that initially, GDP explains most of

the share and second contribution is CO2 emissions in short run. When we move to long run,

GDP contribution starts decreasing slowly while CO2 emissions contribution is increasing in

the explanation of GDP square. Impulse response function shows that GDP is positively

explaining CO2 emissions while GDPSQ has insignificant relation with CO2 emissions in short

run (from period 1-8). In long run, GDP has positive effect on CO2 emissions and GDPSQ has

negative influence on CO2 emissions that is clear from second and third image in first row (Fig.

4). These are similar conclusions drawn from VECM for causality and ARDL for EKC. Thus,

our variables behave on similar pattern even out of sample.

4. Conclusion

This paper investigated long run and causal relation between economic growth and carbon

dioxide emissions based on Environmental Kuznets Curve for Croatia. Quarterly data has been

used for the period of 1992Q1-2011Q1. In order to test the validity of EKC, ARDL bounds test
 ACCEPTED MANUSCRIPT

approach has been used to detect cointegration among the variables and if cointegration exists,

next step was to confirm the validity of EKC. F-statistics results were inconclusive. Therefore,

we have decided cointegration on the basis of negative and significance error correction term.

After confirming long run relation, we move to check short and long run coefficients. GDP

coefficient was having positive sign while GDP square was with negative sign and both

variables were statistically significant in long run. Thus, it confirms the validity of EKC for

Croatia in long run. Negative and highly significance ECT showed the disturbance in the short

run equilibrium will be adjusted in long run with the adjustment speed of 5.3% quarterly. Short

run results show the absence of EKC as it is long run Phenomena. Different diagnostic tests

confirm the perfectness of model. DOLS and FMOLS results confirm the robustness of long

run results. VECM model shows bidirectional causality between environmental pollution and

economic growth in short run. It shows variables’ interdependency on each other in short run.

There was unidirectional causality running from economic growth to CO2 emissions in long

run. It implies that pollution emissions reduction and investment on pollution reduction will

not hurt the economic growth of Croatia. Further, variance decomposition and impulse

response function yield the similar responses. All results are witness to confirm the validity of

EKC in Croatia in long run.

From the results of this research, it can be concluded that economic growth is supportive to

overcome environmental pollution in Croatia. Pollution will decrease with the growth and

development. Thus, Croatian Government should encourage growth and development to tackle

CO2 emissions. In order to get higher economic growth, dirty and less efficient technologies

should be removed from production units with environmental friendly technologies.

Renewable energy should be introduced in the production process as it will not only maintain

and encourage growth and development but will also be supportive to reduce carbon emissions.

Other factors such as renewable energy, biofuel energy production, advance technology usage
 ACCEPTED MANUSCRIPT

to control CO2 emissions, environmental awareness, urbanization, FDI, employment can be

used as the determinants of carbon dioxide emissions. However, we leave for future research.

Acknowledgements

We would like to thank Editor of the journal, Prof. Dr. Isabel Soares, and three anonymous

reviewers for their constructive and valuable comments and suggestions in preparing this final

version of the manuscript. Authors are also thankful to Katarina Kostelic, Assistant Professor

and project research expert at University of Juraj Dobrila of Pula, Croaita for her supportive

role in sharing reading materials and valuable discussion related to this manuscript. Thanks to

Dr. Xu Yan, Dongbei University of Finance and Economics, Dalian, China for econometrics

discussion. Of course, all the errors and mistakes remain our own.

Funding: This paper is financially supported by the Major Project “The Key and Difficult

point of Changing Economic Growth Mode: Risk Analysis, Control System and Incentive

Mechanism” Funded by National Social Science Fund, China with Grant No. 12&ZD076”.

Any opinions, findings and conclusions or recommendations expressed in this paper are those

of authors and don’t necessarily reflect the views of funding body.

References
[1] Fodha, M., Zaghdoud, O., 2009. Economic growth and pollutant emissions in Tunisia: an
empirical analysis of the environmental Kuznets curve. Energy Policy 38, 1150–1156.

[2] Baek, J., & Pride, D., 2014. On the income-nuclear energy-CO2 emissions nexus
revisited. Energy Economics 43, 6–10.

[3] Andreoni, V., & Galmarini, S. ,2016. Drivers in CO2 emissions variation: A
decomposition analysis for 33 world countries. Energy 103, 27–37.

[4] Lau, L.S., Choong, C.K., Eng, Y.K., 2014. Investigation of the environmental Kuznets
curve for carbon emissions in Malaysia: do foreign direct investment and trade matter?
Energy Policy 68, 490–497.
[5] Tsai, B.-H., Chang, C.-J., & Chang, C.-H. ,2016. Elucidating the consumption and CO2
emissions of fossil fuels and low-carbon energy in the United States using Lotka–Volterra
models. Energy 100, 416–424.
 ACCEPTED MANUSCRIPT

[6] Stern N. The economics of climate change: the Stern review. UK: Cambridge University
Press; 2007.

[7] IPCC (2014). Climate Change 2014- Mitigation for Climate Change, 5th Assessment
Report .Intergovernmental Panel on Climate Change, Working Group III. Cambridge
University Press [Edenhofer, O., R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K.
Seyboth, A. Adler, I. Baum, S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S.
Schlömer, C. von Stechow, T. Zwickel and J.C. Minx (eds.)], Cambridge and New York.
http://mitigation2014.org/report/publication/
[8] IEA, I.E.A. World energy outlook. Paris: OECD/IEA; 2008.
[9] Martinho, V. J. P. D. , 2016. Energy consumption across European Union farms: Efficiency
in terms of farming output and utilized agricultural area. Energy 103, 543–556.
[10] Rahman, S., Hanifa, A., Shahari, F., Aslam, M., Sok, C., Ruhana, C., & Pervin, S., 2016.
Ef fi cient energy consumption in industrial sectors and its effect on environment : A
comparative analysis between G8 and Southeast Asian emerging economies. Energy, 97, 82–
89.
[11] Rothman, D.S., de Bruyn, S.M., 1998. Probing into the environmental Kuznets curve
hypothesis. Ecological Economics 25, 143–145.

[12] Kuznets, S., 1955. Economic growth and income inequality. American Economic
Review 45, 1–28.

[13] Grossman, G.M., Krueger, A.B., 1991. Environmental Impacts of a North American
free trade agreement. NBER Working Paper 3914. NBER, Cambridge.

[14] Acaravci, A., Ozturk, I., 2010. On the relationship between energy consumption, CO2
emissions and economic growth in Europe. Energy 35, 5412–5420.
[15] Akbostanci, E., Turut-Asik, S., Tunc, I.G., 2009. The relationship between income and
environment in Turkey: is there an environmental Kuznets Curve? Energy Policy 37 (2),
861–867.
[16] Apergis, N., Payne, J.E., 2010. The emissions, energy consumption, and growth nexus:
Evidence from the commonwealth of independent states. Energy Policy. 38, 650–655.
[17] Jaunky, V.C., 2011. The CO2 emissions-income nexus: evidence from rich countries.
Energy Policy 39, 1228–1240.
[18] Lotfalipour, M.R., Falahi, M.A., Ashena, M., 2010. Economic growth, CO2 emissions,
and fossil fuels consumption in Iran. Energy 35, 5115–5120.
[19] Nasir, M., Rehman, F., 2011. Environmental Kuznets Curve for carbon emissions in
Pakistan: an empirical investigation. Energy Policy 39, 1857–1864.
[20] Ozturk, I., Acaravci, A., 2010. CO2 emissions, energy consumption and economic
growth in Turkey. Renew. Sustain. Energy Rev. 14 (9), 3220–3225.
[21] Ozturk, I., Acaravci, A., 2013. The long-run and causal analysis of energy, growth,
openness and financial development on carbon emissions in Turkey. Energy Econ. 36, 262–
267.
 ACCEPTED MANUSCRIPT

[22] Pao, H., Tsai, C., 2011. Modeling and forecasting the CO2 emissions, energy
consumption, and economic growth in Brazil. Energy 36, 2450–2458.
[23] Pao, H.-T., Yu, H.-C., & Yang, Y.-H., 2011. Modeling the CO2 emissions, energy use,
and economic growth in Russia. Energy, 36(8), 5094–5100.
[24] Saboori, B., Sulaiman, J., Mohd, S., 2012. Economic growth and CO2 emissions in
Malaysia: a cointegration analysis of the Environmental Kuznets Curve. Energy Policy 51,
184–191.
[25] Saboori, B., Sulaiman, J., 2013a. CO2 emissions, energy consumption and economic
growth in Association of Southeast Asian Nations (ASEAN) countries: a co-integration
approach. Energy 55, 813–822.
[26] Saboori, B., Sulaiman, J., 2013b. Environmental degradation, economic growth and
energy consumption: evidence of the environmental Kuznets curve in Malaysia. Energy
Policy 60, 892–905.
[27] Shahbaz, M., Khraief, N., Uddin, G.S., Ozturk, I., 2014. Environmental Kuznets curve
in an open economy: a bounds testing and causality analysis for Tunisia. Renew. Sustain.
Energy Rev. 34, 325–336.
[28] Shahbaz, M., Ozturk, I., Afza, T., Ali, A., 2013. Revisiting the environmental Kuznets
curve in a global economy. Renew. Sustain. Energy Rev. 25 (2013), 494–502.
[29] Shahbaz, M., Solarin, S.A., Mahmood, H., Arouri, M., 2013. Does financial devel-
opment reduce CO2 emissions in Malaysian economy? A time series analysis. Econ. Model.
35, 145–152.
[30] Shahbaz, M., Mutascu, M., Azim, P., 2013. Environmental Kuznets curve in Romania
and the role of energy consumption. Renew. Sustain. Energy Rev. 18, 165–173.
[31] Wang, S.S., Zhou, D.Q., Zhou, P., Wang, Q.W., 2011. CO2 emissions, energy
consumption and economic growth in China: a panel data analysis. Energy Policy 39, 4870–
4875.
[32] Yavuz, N.Ç, 2014. CO2 emission, energy consumption, and economic growth for
Turkey: evidence from a cointegration test with a structural break. Energy Sources Part B:
Econ. Plan. Policy 9 (3), 229–235.

[33] Xu, B., & Lin, B., 2016. Reducing the CO2 emissions in China's manufacturing
industry: Evidence from nonparametric additive regression models. Energy 101, 161–173.

[34] Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approaches to the analysis of
level relationships. J. Appl. Econ. 16, 289–326.
[35] Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction representation:
estimation and testing. Econometrica 55, 251–276.
[36] Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on
cointegration with applications to the demand for money. Oxf. Bull. Econ. Stat. 52, 169–210.
[37] Phillips,P.C.B.,Hansen,B.E.,1990.Statistical Inference in Instrumental variables
regression with I(1) process.Rev.Econ.Stud.57,99–125.
[38] Kao, C., Chiang, M.H., 2000. On the estimation and inference of a cointegrated regression
 ACCEPTED MANUSCRIPT

in panel data. Center for Policy Research, Paper 145.

[39] Stock, J.,Watson, M.W.,1993. A simple estimator of cointegrating vector in higher order
integrated systems.Econometrica 6,783–820.
[40] Payne, J.E., 2002. Inflationary dynamics of a transitional economy: the Croation
experience. Journal of Policy Modeling 34, 219–230.
[41] Lütkepohl, Helmut, 1991. Introduction to Multiple Time Series Analysis, New York:
Springer-Verlag.
[42] Tiwari, A.K., Shahbaz, M., Hye, Q.M.A., 2013. The environmental Kuznets curve and
the role of coal consumption in India: cointegration and causality analysis in an open
economy. Renew. Sustain. Energy Rev. 18, 519–527.
[43] Stern, D.I., Common, M.S., Barbier, E.B., 1996. Economic growth and environmental
degradation: the environmental Kuznets curve and sustainable development.World Dev. 24,
1151–1160.
[44] The needed Development of New Power Plants and Facilities in the Republic of Croatia
in the Period 2001-2020 (Master Plan) Energetski Institut Hrvoje Po.ar 2001.
[45] Grani, G., Pešut, D., Zeljko, M., Bošnjak, R., Kulišiæ, B., Tot, M., Fabek, R., 2011.
What processes can be expected in the Croatian energy sector by 2050. Nafta Scientific
Journal, 62(3-4), 87–95.

[46] Esteve, V., Tamarit, C., 2012. Threshold cointegration and nonlinear adjustment
between CO2 and income: the environmental Kuznets Curve in Spain, 1857– 2007. Energy
Econ. 34, 2148–2156.
[47] Iwata, H., Okada, K., Samreth, S., 2010. Empirical study on the environmental Kuznets
curve for CO2 in France: the role of nuclear energy. Energy Policy 38, 4057–4063.
[48] Jalil, A., Feridun, M., 2011. The impact of growth, energy and financial development on
the environment in China: a cointegration analysis. Energy Econ. 33, 284–291.
[49] Du, L., Wei, Chu, Cai, S., 2012. Economic development and carbon dioxide emissions
in China: provincial panel data analysis. China Econ. Rev. 23, 371–384.
[50] Pesaran, H.M., Shin, Y., 1999. Autoregressive distributed lag modelling approach to
cointegration analysis. In: Storm, S. (Ed.), Econometrics and Economic Theory in the 20th
Century: The Ragnar Frisch Centennial Symposium. Cambridge University Press,
Cambridge.
[51] Chen, J., & Huang, Y., 2013. The Study of the Relationship between Carbon Dioxide (
CO2 ) Emission and Economic Growth, 6(December), 45–61.
[52] Borensztein, E., De Gregorio, J., & Lee, J.-W., 1998. How does foreign direct
investment affect economic growth? Journal of International Economics, 45(1), 115–135.
[53] Banerjee, A., Dolado, J., and Mestre, R., 1998. Error‐Correction Mechanism Tests for
Cointegration in A Single‐Equation Framework. Journal of Time Series Analysis, 19(3):
267-283.
[54] Boutabba, Mohamed Amine. 2014. The Impact of Financial Development, Income,
Energy and Trade on Carbon Emissions: Evidence from the Indian Economy. Economic
Modelling 40. Elsevier B.V.: 33–41.
 ACCEPTED MANUSCRIPT

[55] Granger, G.W.J., 1969. Investigating causal relations by econometric model and cross
spectral methods. Econometrica 37, 424–438.
[56] Enders W. Applied econometric time series. New York: Wiley; 2015.
[57] Banerjee, A., 1999. Panel data unit roots and cointegration: an overview. Oxford
Bulletin of Economics and Statistics 61 (S1), 607–629.
[58] Ouattara, B., 2004. The Impact of Project Aid and Programme Aid on Domestic
Savings: A Case Study of Côte d'Ivoire. Centre for the Study of African Economies
Conference on Growth, Poverty Reduction and Human Development in Africa, April
(2004).

[59] Zaman, K., Shahbaz, M., Loganathan, N., & Ali, S., 2016. Tourism development ,
energy consumption and Environmental Kuznets Curve : Trivariate analysis in the panel of
developed and developing countries. Tourism Management 54, 275–283.

[60] Sohag, K., Ara, R., Mastura, S., & Abdullah, S., 2015. Dynamics of energy use ,
technological innovation , economic growth and trade openness in Malaysia. Energy 90,
1497–1507.

[61] Halicioglu, F., & Ketenci, N., 2016. The impact of international trade on environmental
quality : The case of transition countries. Energy 109, 1130–1138.

[62] Rahman, S., Hanifa, A., Shahari, F., Aslam, M., Sok, C., Ruhana, C., & Pervin, S., 2016.
Efficient energy consumption in industrial sectors and its effect on environment : A
comparative analysis between G8 and Southeast Asian emerging economies. Energy 97, 82–
89.

[63] Hao, Y., Liu, Y., Weng, J., & Gao, Y., 2016. Does the Environmental Kuznets Curve for
coal consumption in China exist ? New evidence from spatial econometric analysis. Energy
114, 1214–1223.

[64] Bahmani-Oskooee, M., & Brooks, T. J., 2003. A new criteria for selecting the optimum
lags in Johansen’s cointegration technique. Applied Economics, 35, 875.
[65] Lutkepohl H. Structural vector autoregressive analysis for cointegrated variables. AStA
Advances in Statistical Analysis 2006; 90:75–88.

[66] Dinda, S., 2004. Environmental Kuznets curve hypothesis: a survey. Ecological
Economics 49, 431–455.

[67] Al-Mulali, U., Saboori, B., & Ozturk, I., 2015. Investigating the environmental Kuznets
curve hypothesis in Vietnam. Energy Policy 76, 123–131.

[68] Grossmann, G. M., and Krueger, A. B. 1993. Environmental impact of a North

American free trade agreement. In: The Mexico-US Free Trade Agreement, Garber, P. (Ed.).
Cambridge, MA: MIT Press, pp. 13–56.

[69] Grossmann, G. M., and Krueger, A. B. 1995. Economic growth and the environment. Q.
J. Econ. 110:353–377.
 ACCEPTED MANUSCRIPT

[70] List, J.A., Gallet, C.A., 1999. The Environmental Kuznets Curve: dose one size fit all?
Ecological Economics 31, 409–423.

[71] Alvarez, M. T. G.-, Moreno, B., & Soares, I., 2016. Analyzing the environmental and
resource pressures from European energy activity : A comparative study of EU member
states. Energy 115, 1375-1384.

Table 1
Results of F-test for the existence of cointegration
Dep.Var F-Statistics Decision
FCO2(CO2/GDP, GDPSQ) 3.18 Inconclusive
FGDP(GDP/CO2, GDPSQ) 3.17 Inconclusive
FGDPSQ(GDPSQ/CO2, GDP) 2.91 Inconclusive
Note: The critical value ranges of f-statistics are 4.13-5.00, 3.10-3.87 and 2.63-3.35 at 1%, 5% and 10% level of
significance, respectively.

Table 2
Short run and long run estimation results
Variable Coefficient
Long Run
GDP 6.465**
(0.031)
GDPSQ -0.832**
(0.044)
C -2.558
(0.628)
Short run
∆CO2(-1) 0.670*
(0.000)
∆GDP -1.673
(0.253)
∆GDP(-1) 0.212
(0.886)

∆GDPSQ 0.285
(0.165)
∆GDPSQ(-1) -0.051
(0.806)
ECTt-1 -0.053*
(0.008)
DW-Statistics 1.90
Note: P-values in parentheses. * and ** represent 1%

and 5% level of significance, respectively

Fig.1. Long Run Environmental Kuznets Curve Shape

 ACCEPTED MANUSCRIPT

Table 3

Diagnostic tests for CO2 emissions equation

Diagnostic tests Null Hypothesis F-statistic Decision

Breusch Godfrey serial Correlation H0: No serial correlation 0.26(0.77) Don’t reject H0
LM test
Heteroskedasticity Test: Breusch- H0:Homoskedasticity 0.63(0.75) Don’t Reject H0
Pagan-Godfrey
ARCH test H0:Homoskedasticity 0.005(0.95) Don’t Reject H0
Normality H0: Residuals are normally 1.98(0.24) Don’t Reject H0
distributed
Ramsey Reset test H0: model is correctly specified 0.04(0.84) Don’t Reject H0
Note: P-values in parentheses. Source: Authors’ estimations using Eviews 9.0

Fig. 2. Plot of Cumulative Sum of Recursive Residuals

 ACCEPTED MANUSCRIPT

30

20

10

-10

-20

-30
94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11

CUSUM 5% Significance

Fig. 3. Plot of Cumulative Sum of Squares of Recursive Residuals

1.2

1.0

0.8

0.6

0.4

0.2

0.0

-0.2
94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11

CUSUM of Squares 5% Significance

Table 4
Robustness Check for long run coefficients

Variables DOLS FMOLS

GDP 5.11* 5.13*
(0.002) (0.001)
GDPSQ -0.63* -0.63*
(0.006) (0.004)
Adj. R2 0.93 0.88
Note: P-values in parentheses. * represents 1% level of significance.

Table 5
Granger Causality Results
 ACCEPTED MANUSCRIPT

Dependent Sources of causation

Variables Short run Long run
∆CO2 ∆GDP ∆GDPSQ ECTt-1
∆CO2 - 3.67* -0.46* -0.68*
(0.00) (0.00) (0.00)
∆GDP 0.047* - 0.13* 0.98*
(0.00) (0.00) (0.00)
∆GDPSQ -0.32* 7.43* - 0.99*
(0.00) (0.00) (0.00)
Note: P-values in parentheses are from f-test for short run and t-test for long run. * and ** represent 1% and 5% level of
significance, respectively

Table 6
Variance Decomposition Method
CO2 emissions GDP GDP2
Period
CO2 Y Y2 CO2 Y Y2 CO2 Y Y2
1 100.00 0.00 0.00 14.41 85.59 0.00 15.22 84.26 0.52
2 99.34 0.51 0.16 15.10 84.87 0.03 15.92 83.33 0.75
3 97.89 1.73 0.38 15.91 84.00 0.09 16.73 82.29 0.99
4 95.90 3.56 0.54 16.82 83.00 0.19 17.61 81.16 1.23
5 93.52 5.89 0.59 17.80 81.90 0.30 18.56 79.98 1.46
6 90.85 8.61 0.54 18.83 80.74 0.43 19.55 78.78 1.68
7 87.98 11.58 0.44 19.89 79.55 0.56 20.54 77.59 1.86
8 84.96 14.67 0.37 20.94 78.38 0.68 21.53 76.46 2.01
9 81.84 17.75 0.41 21.96 77.25 0.79 22.48 75.40 2.12
10 78.70 20.68 0.62 22.94 76.19 0.87 23.39 74.42 2.19
11 75.60 23.36 1.04 23.86 75.22 0.92 24.24 73.55 2.21
12 72.58 25.71 1.71 24.71 74.34 0.95 25.03 72.78 2.19
13 69.72 27.69 2.59 25.49 73.56 0.95 25.75 72.12 2.13
14 67.05 29.29 3.66 26.19 72.88 0.93 26.40 71.56 2.05
15 64.61 30.51 4.87 26.82 72.29 0.89 26.98 71.08 1.94
16 62.42 31.41 6.17 27.37 71.79 0.84 27.49 70.68 1.84
17 60.49 32.02 7.49 27.85 71.36 0.79 27.93 70.34 1.73
18 58.81 32.39 8.80 28.26 71.00 0.75 28.31 70.06 1.63
19 57.36 32.58 10.06 28.60 70.68 0.71 28.63 69.82 1.55
20 56.13 32.63 11.23 28.90 70.40 0.70 28.91 69.60 1.49

Fig.4.

Impulse Response Function (IRF)

 ACCEPTED MANUSCRIPT

Response of CO2 to CO2 Response of CO2 to GDP Response of CO2 to GDPSQ

.03 .03 .03

.02 .02 .02

.01 .01 .01

.00 .00 .00

-.01 -.01 -.01

-.02 -.02 -.02

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

Response of GDP to GDP Response of GDP to CO2 Response of GDPSQ to CO2

.04 .04 .3

.03 .03
.2

.02 .02
.1

.01 .01
.0
.00 .00

-.1
-.01 -.01

-.02 -.02 -.2

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

APPENDIX
Summary of previous research on Environmental Kuznets Curve (EKC)
Study Country Period Technique Conclusion
ARDL bound tests, EKC valid in short and
Lau et al. (2014) Malaysia 1970-2008
VECM long run
ARDL bound tests, EKC valid in short and
Shahbaz et al. (2014) Tunisia 1971-2010
VECM long run
Akbostanci, Türüt- Johansen EKC is not valid in short
Turkey 1968-2003
Aşik, & Tunç (2009) cointegration and long run
11 countries of the
Apergis & Payne Panel cointegartion,
Commonwealth of 1992–2004 EKC valid in long run
(2010) VECM, FMOLS
Independent States
36 high income Not valid in short and long
Jaunky (2011) 1980–2005 GMM, VECM
countries run
Johansen EKC valid in long run but
Nasir & Rehman (2011) Pakistan 1972–2008
cointegration,VECM invalid in short run
Wang, Zhou, Zhou, & Panel Co-integration,
China 1995–2007 EKC is not valid in China
Wang (2011) VECM
EKC valid for short and
Saboori, Sulaiman, & long run in ARDL but in
Malaysia 1980-2009 ARDL, VECM
Mohd (2012) VECM short run results
fail to confirm EKC
Saboori & Sulaiman ARDL, Johansen Co-
Malaysia 1980–2009 Mix results
(2013) integration, VECM
 ACCEPTED MANUSCRIPT

Johansen co-
Pao & Tsai (2011) Brazil 1980-2007 EKC valid in long run
integration, VECM
Johansen Co- EKC is not valid in short
Pao et al. (2011) Russia 1990-2007
integration and long run
Ozturk & Acaravci EKC not valid in short and
Turkey 1968–2005 ARDL, VECM
(2010) long run
EKC valid in short and
Ozturk & Acaravci
Turkey 1960–2007 ARDL, VECM long run. However VECM
(2013)
yields opposite results.
Different
Acaravci & Ozturk 19 European EKC valid for only two
years for each ARDL, VECM
(2013) countries countries.
country
EKC valid in short and
Shahbaz et al.(2013) Romania 1980–2010 ARDL
long run
EKC Valid in short and
Tiwari et al.(2013) India 1966–2009 ARDL, VECM
long run
EKC valid for two
Saboori & Sulaiman Five ASEAN
1971-2009 ARDL, VECM countries in the long run
(2013b) countries
only
Johansen co-
Yavuz (2014) Turkey 1960-2007 EKC valid in long run
integration
Al-Mulali, Saboori, & EKC is invalid in short
Vietnam 1981–2011 ARDL
Ozturk (2015) and long run
Nonparametric
Xu and Lin (2016) China 2000-2013 EKC valid in China
additive regression
three diversified
Zaman et al. (2016) 2005-2013 two stage least square Validity of EKC
region of the world
Hao et al. (2016) China 1995-2012 Spatial Durbin Model EKC is valid in China
Halicioglu and Ketenci 15 transitions EKC for only three
1991-2013 ARDL and GMM
(2016) countries countries