How effective are energy ef

ﬁciency and renewable energy in curbing

emissions in the long run? A heterogeneous panel data analysis

Fatih Cemil €

Ozbu

gday

, Bahar Celikkol Erbas

Department of Economics, Y

ıldırım Beyazıt University, Cinnah Cad. No: 16, Çankaya, Ankara, Turkey

Department of Economics, TOBB University of Economics and Technology, Sogutozu Cad. No: 43, 06530 Sogutozu, Ankara, Turkey

a r t i c l e i n f o

Article history:
Received 19 June 2014
Received in revised form
6 January 2015
Accepted 27 January 2015
Available online 4 March 2015

Keywords:
Global warming
CO

emissions

Energy ef

ﬁciency

Renewable resources
Industrialization
Common correlated effects estimator model

a b s t r a c t

Energy ef

ﬁciency and renewable energy are considered to be two indispensable solutions to control GHG

(greenhouse gas) emissions. Moreover, industrialization is at the center of discussions on the roles of
countries to reduce CO

emissions. However, the literature is underprovided to understand the long run

contribution of energy ef

ﬁciency, renewable energy and industrial composition in reducing GHG emis-

sions at the macro level. In this study, we differentiate the effects of economic activity, energy ef

ﬁciency,

economic structure and use of renewable energy resources on CO

emissions. We develop energy ef

ﬁ-

ciency indices for thirty six countries for the period of 1971

e2009 and use a CCE (common correlated

effects) estimator model that is consistent under heterogeneity and cross-sectional dependence. We

ﬁnd

a positive signi

ﬁcant effect of energy efﬁciency on CO

emissions in the long-run. Similarly, substituting

renewable energy for non-renewable energy reduces CO

emissions in the long-run. Our results ensure

widely discussed roles of energy ef

ﬁciency and renewable energy in curbing CO

emissions. Furthermore,

the scale of economic activity measured by real income and industrialization have signi

ﬁcant positive

effect on CO

emissions.

1. Introduction

The relationship between CO

emissions and economic activity

is important in understanding global climate change and control-
ling GHG (greenhouse gas) emissions . This relationship is widely
explored through the EKC (Environmental Kuznets Curve) hy-
pothesis, and granger causality and panel cointegration analyses on
the factors affecting CO

emissions in the literature. Dinda (2004)

[13]

and Al-mulali (2012)

[2]

provide extensive reviews on EKC, and

granger causality and panel cointegration analyses, respectively.
While the literature is extensive, energy ef

ﬁciency and renewable

energy, which are considered to be two indispensable solutions to
control GHG (greenhouse gas) emissions, have not been studied as
the factors affecting CO

emissions. In addition to energy ef

ﬁciency

and renewable energy, industrialization is also at the center of
discussions on the roles of countries to reduce CO

emissions. The

net effects of these factors can be best observed over a longer time
horizon. Therefore, we investigate the long-run equilibrium rela-
tionship between these factors and CO

emissions.

We separate economic activity into its components: scale (scale

is sometimes called economic activity), energy ef

ﬁciency, economic

structure and the use of renewable energy resources. The scale
component captures the effect of increases in GDP (gross domestic
product) on CO

emissions keeping economic structure and ef

ﬁ-

ciency constant. Economic structure con

ﬁnes the effects of shifts to

the agriculture, manufacturing and service sectors. Countries may
experience structural changes from pollution-intensive low value
added manufacturing sectors to low polluting high value added
service sectors. The effect of technology on CO

emissions is

embodied in energy ef

ﬁciency. As countries grow, they obtain more

resources that enable the production of energy ef

ﬁcient products.

Greater energy ef

ﬁciency in turn enables higher income with lower

emissions. In the paper, we develop a comprehensive and

reliable measure of energy ef

ﬁciency for each country to study its

effect on emissions and to investigate the potential of energy ef

ﬁ-

ciency policies to reduce emissions. The strategies in use to reduce
CO

emissions suggest that the substitution of renewable energy

resources for non-renewable resources is a well-established way to
reduce emissions, especially the replacement of fossil fuels. While
policy makers suggests energy ef

ﬁciency measures and the use of

renewable energy sources along with other CO

emissions tax or

trading schemes, the literature lacks empirical studies investigating

* Corresponding author. Tel.: þ90 3124667533x3551.

E-mail addresses:

fcozbugday@ybu.edu.tr

(F.C. €

Ozbu

gday),

bcelikkol@etu.edu.tr

(B.C. Erbas).

Contents lists available at

ScienceDirect

Energy

j o u r n a l h o me p a g e :

w w w . e l s e v i e r . c o m/ l o ca t e / e n e r g y

http://dx.doi.org/10.1016/j.energy.2015.01.084

0360-5442/

Energy 82 (2015) 734

e745

the relationship between energy ef

ﬁciency and renewable con-

sumption and CO

emissions using current time series economet-

rics methods with actual data. Therefore, our model speci

ﬁcally

captures the effects of energy ef

ﬁciency via a comprehensive and

reliable measure and the substitution of renewable energy re-
sources for non-renewable sources on CO

emissions enabling us to

ﬁll this gap in the literature and to provide empirical evidence on
the roles of renewable energy and energy ef

ﬁciency in reducing

. To our knowledge, no studies have investigated the effects of

aforementioned economic activities on CO

emissions across

countries and over time.

Finally, we employ the factor decomposition analysis with the

most advanced panel data methodology available, the CCE (com-
mon correlated effects) estimation model of Pesaran (2006)

[32]

investigate the role of energy ef

ﬁciency and renewable energy in

curbing CO

emissions for 36 countries for the period 1971

e2009.

The superiority of the CCE approach arises from the fact that it can
handle heterogeneity and both weak and strong forms of cross-
section dependence. Furthermore, the country-speci

ﬁc effects and

the heterogeneous trend components absorb the effects of any
time-invariant or time-varying omitted variables, thereby elimi-
nating omitted variable bias.

The rest of the paper is organized into four sections. Section

provides a review of the relevant literature. The data are
described, variables are constructed and the estimation method-
ology is introduced in section

. Section

presents the results.

Finally, we provide some concluding remarks in section

2. Literature review

The literature on the relationship between economic growth

and environmental pollution has two main components, the

ﬁrst

focusing on economic growth and environmental pollution (the
economic growth-environmental pollution nexus), and the second
concentrating on economic growth and energy consumption
(economic growth-energy consumption nexus). A voluminous EKC
and Granger causality literature with a bivariate framework has
emerged in studies of both nexuses

[1,3,35,37]

2.1. Economic growth-environmental pollution nexus

The studies in the

ﬁrst part of the literature investigate the

relationship between economic growth and environmental pollu-
tion using income and CO

emissions, the main source of the GHG

effect. Within the

ﬁrst nexus, the increasing number of EKC studies

has motivated the development of two other subsets of studies on
the causal relationship between CO

emissions and income and on

the estimation methodologies used to understand this relationship.

In the EKC portion of the literature, a large number of empirical

studies focus on the shape of the EKC for single countries or groups
of countries. These studies'

ﬁndings regarding the relationship

between CO

emissions and GDP (gross domestic product) vary

from study to study. While some studies verify the inverted-U
shape of the EKC, others

ﬁnd different curve shapes. Based on an

extensive review of the literature, Jaunky (2011)

[19]

discloses that

empirical evidence on the existence of the EKC has always been
mixed and con

ﬂicting. Jaunky (2011)

[19]

speci

ﬁcally emphasizes

that the implementation of renewable energy resources should
accompany direct measures curbing CO

emissions.

Similarly, Piaggio and Padilla (2012)

[34]

study the relationship

between CO

emissions and economic activity for 31 countries for

the period from 1950 to 2006 using a co-integration analysis. They
suggest considering the differences among countries in the rela-
tionship between CO

emissions and economic activity to avoid

faulty estimations and conclusions. They state that differences

would depend on the real determinants of the relationship, which
would be shaped by energy and environmental policies. As inferred
by Piaggio and Padilla (2012)

[34]

and Jaunky (2011)

[19]

; our

model captures the effects of these policies by including energy
ef

ﬁciency and renewable energy resources in reference to non-

renewable resources in a multivariate setting.

Facilitated by EKC studies, studies focusing on the causality

between CO

emissions and income investigate the existence and

the direction of the relationship. Similar to the EKC research, the
empirical evidence for the causal relationship differs from study to
study. These puzzling results on EKC and the causal relationship
between CO

emissions and GDP have motivated researchers to

scrutinize the econometric techniques employed in the relevant
literature. Methodological studies have been carried out to test the
validity of the econometric models, tools and functional forms

[39,40]

. As the estimation and methodology subset of the literature

expands with the desire to eliminate estimation problems and

ﬁnd

appropriate robust techniques, the literature also values studies
that extend the

ﬁndings of existing studies by implementing new

econometric techniques developed by Pesaran (2006)

[32]

for the

panel data

[6,38]

2.2. Economic growth-energy consumption nexus

The second part of the literature, the economic growth-energy

consumption, extensively studies the causal relationship between
energy consumption and GDP. Ozturk (2010)

[26]

provides a

comprehensive review of the literature and concludes that there
are con

ﬂicting results on the existence and the direction of this

causality. Different data sets, country characteristics, variables used
and econometric methodologies are listed as the main reasons for
these con

ﬂicting results. Ozturk (2010)

[26]

suggests that re-

searchers working in this area should employ new approaches and
perspectives rather than simply analyzing different countries and
different time intervals using the same models and methods.

Although many studies focus on bivariate models of energy and

output, a few studies include a third variable such as urbanization,
employment, energy prices, capital or labor in addition to energy
and output (e.g. Refs.

[14]

[21,23,26,38]

). Among these studies, Lee

and Chang (2008)

[21]

touch on the importance of energy ef

ﬁciency

in reducing GHG emissions, however, they do not directly study the
relationship between energy ef

ﬁciency and GHG emissions. They

argue that governments in Asian countries should aim to imple-
ment energy ef

ﬁcient industrial processes to better control GHG

emissions. Given the importance of energy ef

ﬁciency emphasized

in the literature, we incorporate energy ef

ﬁciency in our model to

see its impact on CO

emissions.

Energy intensity is one of the components derived from CO

emissions. Although Wing (2008)

[41]

only studies a single county,

this work is of interest because it focuses on energy intensity in the
U.S. at an aggregate level and the multiple factors behind it over the
period from 1958 to 2000. Speci

ﬁc attention is devoted to technical

change and industry composition. Wing (2008)

[41]

ﬁnds that

changes in industrial composition and disembodied technological
progress are the main sources of the decline in the energy intensity.
Input substitution due to changes in the relative prices of inputs has
only transitory impact. Similarly, Sadorsky's (2014)

[38]

work on the

effect of urbanization on CO

emissions is also of interest since it

captures the effect of technology through energy intensity in
emerging economies. Sadorsky (2014)

[38]

stresses the role of

reduction in energy intensity in reducing CO

emissions in emerging

countries where population and af

ﬂuence are likely to rise.

The other important factor affecting CO

emissions is industri-

alization or changes in the composition of a country's economic
activities. Although Cherniwchan (2012)

[9]

does not focus on CO

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

735

by taking SO

as a proxy for environmental pollution, he demon-

strates that the process of industrialization is a signi

ﬁcant deter-

minant of observed changes in SO

. Mentioned in the previous

paragraph, another study about the role of industrial composition
on CO

emission is Wing (2008)

[41]

; which

ﬁnds that changes in

industrial composition is one of the main sources of the decline in
the energy intensity.

The use of renewable energy is universally seen as a route to

achieve emissions reductions to address the threat of global
warming and climate change in the second nexus. However, the
literature is still de

ﬁcient in empirical studies that examine the

relationship between renewable energy consumption and CO

emissions using up to date time series econometrics. Apergis et al.
(2010)

[5]

and Chiu and Chang (2009)

[10]

are the only two studies

in the literature that investigate the relationship between renew-
able energy consumption and CO

emissions. Apergis et al. (2010)

[5]

ﬁnd that nuclear energy consumption contributes to a reduction

in CO

emissions whereas renewable energy consumption does not

play an important role in reducing CO

emissions. They provide

intermittent supply problems and inadequate storage technology
as the possible reasons for their empirical

ﬁndings on renewable

energy consumption. However, they

ﬁnd bi-directional causality

between renewable energy consumption and economic growth. On
the other hand, Chiu and Chang (2009)

[10]

observe an effect of

renewable energy supply on CO

emission reductions after a certain

threshold and suggest an increase in the proportion of renewable
energy supply above this threshold level to solve the dilemma
between economic growth and CO

emissions. The effect of the use

of renewable energy on CO

can be better understood by consid-

ering the prominence of non-renewable energy and nuclear energy
within a country's energy consumption. Moreover, an economy's
structure and its ef

ﬁciency of energy use are other important fac-

tors affecting CO

emissions. Therefore, differently from Apergis

et al. (2010)

[5]

and Chiu and Chang (2009)

[10]

; we integrate these

variables in our modeling.

This paper studies the effects of GDP, energy ef

ﬁciency, industry

composition and renewable energy at the aggregate level on the
generation of CO

emissions by combining the two nexuses which

is relatively new approach to research in this area and by using of
one of the most recent estimation methodologies that takes the
cross-sectional dependence and heterogeneity of panel data into
account.

3. Empirical strategy and data

There are considerable differences across countries in terms of

income, technological development, resource abundance and en-
ergy policies. This heterogeneity should also appear in CO

emis-

sions and in the responsiveness of CO

emissions to changes in

these variables. CO

emissions can also vary between countries not

only due to veri

ﬁable measures and policies but also to other un-

observed idiosyncratic factors. However, there are considerable
interdependencies in CO

emissions, income and energy policies

across countries. For instance, by refusing to extract extra amounts
of oil, a resource-abundant country might in

ﬂuence world oil pri-

ces, which would impinge on the status of economic activities and
CO

emissions in other countries. Alternatively, a country suffering

from a local economic crisis might change its trade composition
with its trading partners. These shocks, which induce correlations
between pairs of countries, often cannot be observed by econo-
metricians as they affect the economic system as a whole. More
importantly, the effects of such common external shocks could also
differ across countries.

In addition to global shocks such as energy or economic crises,

another source of inter-country dependence is spatial correlation.

Neighboring countries may share common resource-abundance or
economic characteristics that affect income and CO

emissions.

Furthermore, spatial correlation might also be induced by cross-
border CO

spillovers between neighboring countries.

In the presence of heterogeneity and cross-sectional depen-

dence as in the current study, conventional estimators such as
panel data estimators or ordinary least squares estimators are no
longer valid in econometric terms.

To address the problems

generated by the presence of heterogeneity and cross-sectional
dependence, we employ the CCE (common correlated effects)
estimator of Pesaran (2006)

[32]

; which is consistent under het-

erogeneity and cross-sectional dependence

[16]

. Furthermore, the

CCE estimator can accommodate stationary and nonstationary
variables and, cointegration or not, which is an important point
considering that our data's time dimension consists of long periods.
This estimator can be obtained by augmenting a standard Mean
Group or Pooled Fixed Effects regression with the cross-sectional
averages of the dependent variable and explanatory variables in
order to account for cross-sectional dependence.

In brief, the CCE approach can handle heterogeneity and weak

forms of cross-sectional dependence arising from local CO

spill-

over effects across neighboring countries, and strong forms of
cross-sectional dependence such as global oil shocks. Furthermore,
the country-speci

ﬁc effects and the heterogeneous trend compo-

nents absorb the effects of any omitted time-invariant or time-
varying variables

[7]

3.1. The econometric model

The most fundamental critique to conventional bivariate rela-

tionship between income and environment is lack of explicit policy
considerations such as energy ef

ﬁciency and renewable energy.

Panatoyou (1997)

[27]

says

“… while parsimonious in data and

estimation complexity, the conventional approach is basically a
“black box”: it hides more than it reveals since income level is used
as a catch-all surrogate variable for all the changes that take place
with economic development.

” Therefore, by using a factor

decomposition analysis, we build the model below to account for
energy ef

ﬁciency and renewable energy. Factor decomposition

analysis separately captures the effect of economic activity (in-
come), economic structure (manufacturing share) and technology
(energy ef

ﬁciency and renewable energy use). The decomposition

enables us to consider the impacts of two important policy in-
terventions (energy ef

ﬁciency and use of renewable energy) on CO

emission reductions and thus in turn provides us better under-
standing of income and environment relationship as well as the
effectiveness of these policy tools in combating climate chance.

To determine the long-term relationships among energy ef

ﬁ-

ciency, renewable energy, income, sectoral composition of GDP and
CO

emissions, we employ the following simple linear heteroge-

neous panel data model:

;it

¼ a

þ b

ðGDP

Þ þ b

ðPOPULATION

þ b

ðEFFICIENCY

þ b

ðMANUFACTURING SHAREÞ

þ b

ðRENEWABLE=TOTAL ENERGY Þ

þ u

¼ 1; ::::; N; t ¼ 1; …; T

(1)

See, for example, Andrews (2005)

[4]

for a discussion of cross-section regres-

sion in the presence of cross-sectional dependence due to common shocks.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

736

where

is the country-speci

ﬁc intercept and u

is the error term.

2,it

denotes carbon dioxide emissions from fuel combustion in

country i in year t expressed in million tons. These values are ob-
tained from the CO

Emissions from Fuel Combustion Database of

the IEA (International Energy Agency). We proxy income by using
GDP

, which is the GDP (gross domestic product) in country i in

year t. The GDP data are obtained from the UNSD (United Nations
Statistics Division), and these

ﬁgures are at 2005 prices and in US

dollars. From the same data source, we also compile the population
data as another potential driver of CO

emissions in a country:

POPULATION

is the population of country i in year t.

Elsewhere, EFFICIENCY

is the ef

ﬁciency index, which is calcu-

lated based on Metcalf (2008)

[22]

and Nillesen et al. (2013)

[25]

The index is obtained following a decomposition analysis in which
an energy intensity index is perfectly decomposed into economic
ef

ﬁciency and activity indexes with no residual. The efﬁciency

component captures the reduced energy consumption per unit of
economic activity. The activity component controls for shifts from
energy-intensive economic activities to non-energy intensive ac-
tivities while ef

ﬁciency remains constant. The changes in the

composition of economic activity occur among four different sec-
tors: industry, services, transportation and residential. As the key
drivers of energy demand in industry, the service and trans-
portation sectors are value added associated with these sectors.
Household consumption expenditure is used as the key driver of
residential energy demand.

Table 1

presents the sectors and their

associated economic activity measures that are used for the
decomposition.

The IEA manages the energy balance databases of OECD

(Organisation for Economic Co-operation and Development) and a
selection of non-OECD countries. Energy consumption data are
gathered from the energy balance databases. Value added and GDP
data at 2005 prices in US dollars are obtained from the UNSD
(United Nations Statistics Division).

Following Metcalf (2008)

[22]

; energy intensity can be mathe-

matically speci

ﬁed as a function of energy efﬁciency and economic

activity:

≡

(2)

where E

is aggregate energy use in year t, Y

is real GDP in year t, E

is energy consumption in sector i in year t, and Y

is a measure of

the economic activity in sector i in year t.

Equation

(2)

denotes the

aggregate energy intensity as a function of sector-speci

ﬁc energy

ﬁciency (e

) and sectoral activity (s

As an initial step, the energy intensity index I

¼ e

, e

constructed by referring to the aggregate energy intensity for a base
year. Metcalf (2008)

[22]

claims that it is possible to decompose an

energy intensity index (I

) into economic ef

ﬁciency (F

eff

) and ac-

tivity indexes (F

act

) with no residual using the Fisher Ideal index:

≡I

¼ F

eff

act

(3)

Building the Fisher Ideal index requires the construction of

Laspeyres and Paasche composition and ef

ﬁciency indexes. Eqs.

(4)

and (5)

provide the Laspeyres indexes and Eqs.

(6) and (7)

provide

the Paasche indexes:

act

(4)

eff
t

(5)

act

(6)

eff

(7)

By employing the Laspeyres and Paasche indexes, the Fisher

Ideal indexes are then given by

act

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
L

act

(8)

eff

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
L

eff
t

(9)

If the ef

ﬁciency index for a speciﬁc country in 1990 is 70, then

the energy intensity in that country would have been 70 percent of
its 1971 level in 1990 if the composition of economic activity had
not changed between 1971 and 2009. In other words, there is 30%
increase in energy ef

ﬁciency between these years. Thus, lower ef-

ﬁciency index values indicate greater efﬁciency. It is expected that
the index have positive coef

ﬁcients indicating the fact that de-

creases in the index, presenting the improvements in energy ef

ﬁ-

ciency, decrease CO

emissions.

Changes

economic

structure

are

captured

(MANUFACTURING

SHARE)

which

denotes

the

output

manufacturing industries as a share of total GDP. To focus on the
effect of industrialization in the presence of changing GDP, we

Table 1
Sectors for decomposition analysis at country level.

Sector

Economic activity

Sectoral energy ef

ﬁciency

Measure

Residential

Household
consumption
expenditures ($2005)

Energy consumption (in thousand
tonnes of oil equivalent (ktoe) on a
net calori

ﬁc value basis) per

dollar ($2005)

Services

Value added in
services sector
($2005)

Energy consumption (in thousand
tonnes of oil equivalent (ktoe) on a
net calori

ﬁc value basis) per

dollar ($2005)

Industrial

Value added in
industrial
sector ($2005)

Energy consumption (in thousand
tonnes of oil equivalent (ktoe) on a
net calori

ﬁc value basis) per

dollar ($2005)

Transportation

Value added in
transportation
sector ($2005)

Energy consumption (in thousand
tonnes of oil equivalent (ktoe) on a
net calori

ﬁc value basis) per

dollar ($2005)

Total

GDP ($2005)

Energy consumption (in thousand
tonnes of oil equivalent (ktoe) on a
net calori

ﬁc value basis) per

dollar ($2005)

Source: Energy and electricity consumption data from the International Energy
Agency. Economic data from the United Nations Statistics Division (UNSD). Data
runs from 1971 to 2009.

POPULATION

is a control variable that is considered to affect CO

emissions.

Energy use in the sectors must sum to aggregate energy use, while the mea-

sures of economic activity do not need to add up to GDP. Furthermore, they do not
necessarily have to be in the same units.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

737

include GDP

and (MANUFACTURING SHARE)

in the model. These

two variables are easy to interpret and encapsulate the effect of the
activity

index

formulated

the

decomposition

analysis.

Conversely, the activity index is not included in the model to avoid
double counting the effect of GDP and the share of manufacturing.
Furthermore, a disadvantage of the activity index is that it only
reveals changes in energy intensity due to overall structural
transformations in the economy without fully explaining the exact
sources of these changes. Finally, the effect of the use of renewable
energy on CO

can be better understood by considering it relative to

the uses of non-renewable and nuclear energy. Therefore, to study
the effect of renewable energy in association with total sources on
CO

emissions, we include (RENEWABLE/TOTAL ENERGY)

, which is

the ratio of renewable energy consumption to total energy con-
sumption (in thousand tons of oil equivalent (ktoe) on a net calo-
ri

ﬁc value basis). The renewables in the data set are solar, wind,

biomass, geothermal, hydro power and ocean resources, solid
biomass, biogas and liquid biofuels as well as biodegradable solid
waste. Countries have their idiosyncratic features in their renew-
able mix. Solid biomass (mainly fuelwood used for cooking in
developing countries) is by far the largest renewable energy source,
representing more than 10% of world TPES (total primary energy
supply), or three-quarters of global renewables supply. Moreover,
as renewables include hydro power, several countries such Brazil,
China and India have large hydro power shares in their energy mix.

After juxtaposing the data from different sources, we obtain a

sample of 1373 observations for 36 countries (

Table 2

). According

to 2009

ﬁgures, these countries, cover 74% of global output, are

responsible for approximately 77% of the world's total CO

emis-

sions. For the great majority of observations, the data run span
from 1971 to 2009. A short description of the variables and the
data sources is provided in

Table 3

Table 4

displays the summary

statistics for our variables. One striking observation is that the
most energy inef

ﬁcient countries on average are oil-abundant

countries such as Saudi Arabia, United Arab Emirates, Iran and
Venezuela. Furthermore, three of these countries (Saudi Arabia,
United Arab Emirates and Venezuela) have the highest average
share of manufacturing industries as a percentage of total GDP,
followed by China. Another stunning observation is that the
average ratios of renewable to total-renewable energy use are
much higher for developing countries than for developed
countries.

Our estimation strategy is as follows: we estimate the model

in [1] using the CCE approach and also present the results from
Pesaran and Smith (1995)'s

[29]

Mean Group (MG) estimator for

comparison purposes. The difference between Pesaran (2006)'s

[32]

CCE approach and Pesaran and Smith (1995)'s

[29]

approach is that the former allows for cross-sectional depen-
dence while the latter does not. Finally, we do not regroup the
countries in our sample based on geographic proximity or based
on the level of development. This is because our estimation
method already incorporates heterogeneity at a more dis-
aggregated level by allowing for a unique coef

ﬁcient for each

country.

Finally, we use a log

elog model that allows for calculation of

elasticities and interpretation in percentage terms.

4. Results

4.1. Preliminary data analysis

We begin our analysis with a preliminary investigation of spatial

dependence at the country level.

Table 5

presents the results of the

cross-sectional dependence tests for variables in both levels and
ﬁrst-differences. The cross-sectional dependence statistics and
associated p-values strongly rejects the null of cross-section inde-
pendence indicates that the cross-correlations are statistically sig-
ni

ﬁcant, implying the existence of cross-sectional correlation

among the countries in our sample. This con

ﬁrms that our choice of

an estimation technique is appropriate.

Table 2
Sample makeup: countries.

Country

Obs.

Coverage

Argentina

2.84%

1971

e2009

Australia

2.77%

1972

e2009

Austria

2.84%

1971

e2009

Belgium

2.84%

1971

e2009

Brazil

2.84%

1971

e2009

Canada

2.77%

1971

e2008

Chile

2.84%

1971

e2009

China, People's Republic of

2.84%

1971

e2009

Denmark

2.84%

1971

e2009

Finland

2.84%

1971

e2009

France

2.84%

1971

e2009

Germany

2.84%

1971

e2009

Hungary

2.84%

1971

e2009

India

2.84%

1971

e2009

Indonesia

2.84%

1971

e2009

Iran, Islamic Republic of

2.69%

1971

e2007

Ireland

2.84%

1971

e2009

Italy

2.84%

1971

e2009

Japan

2.84%

1971

e2009

Mexico

2.84%

1971

e2009

Netherlands

2.84%

1971

e2009

Norway

2.84%

1971

e2009

Portugal

2.84%

1971

e2009

Republic of Korea

2.84%

1971

e2009

Saudi Arabia

2.84%

1971

e2009

South Africa

2.84%

1971

e2009

Spain

2.84%

1971

e2009

Sweden

2.84%

1971

e2009

Switzerland

2.18%

1980

e2009

Tunisia

2.84%

1971

e2009

Turkey

2.84%

1971

e2009

United Arab Emirates

2.55%

1975

e2009

United Kingdom of Great

Britain and Northern Ireland

2.84%

1971

e2009

United States

2.84%

1971

e2009

Uruguay

1.97%

1983

e2009

Venezuela (Bolivarian Republic of)

2.69%

1971

e2007

Total

1373

100.00%

Table 3
Variable de

ﬁnitions and data sources.

Variable

ﬁnition

Source

emissions in million tons

International Energy
Agency (IEA) (CO

emissions from fuel
combustion database)

GDP

GDP in million dollars (2005
prices)

United Nations Statistics
Division (UNSD)

Population

Population in hundred
thousands

United Nations Statistics
Division (UNSD)

ﬁciency

ﬁciency index

Own calculations based on
Metcalf (2007) and Nillesen
et al. (2013)

[25]

Manufacturing

Share of manufacturing
industries in total GDP

World Bank (World
DataBank)

Renewable/total

energy

Ratio of renewable to total
energy

International Energy
Agency (IEA) (Energy
Balances of OECD and
non-OECD Countries
Database)

For more technical details about the estimation procedure see Refs.

[11,30,32]

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

738

Table 4
Summary statistics.

Country

emissions in million tons

GDP in million dollars (2005 prices)

Obs.

Mean

Std. dev.

Min

Max

Obs.

Mean

Std. dev.

Min

Max

a
Argentina

114.54

26.08

174

139,974.40

36,012.05

99,000

230,000

Australia

266.72

78.11

144

395

492,820.50

183,029.30

260,000

850,000

Austria

58.64

7.46

222,051.30

60,139.29

130,000

330,000

Belgium

114.95

8.97

101

133

281,794.90

68,781.57

170,000

400,000

Brazil

221.33

75.14

361

622,051.30

199,844.40

250,000

1,000,000

Canada

452.62

66.09

339

568

768,974.40

244,204.30

400,000

1,200,000

Chile

35.36

17.26

66,025.64

34,684.66

27,000

130,000

China

2625.44

1604.28

800

6832

950,512.80

936,359.00

130,000

3,500,000

Denmark

55.72

5.06

193,076.90

45,546.61

130,000

270,000

Finland

53.97

7.31

136,359.00

40,999.67

75,000

220,000

France

393.00

43.85

341

485

1,592,821.00

385,185.90

940,000

2,200,000

Germany

931.64

92.78

750

1104

2,182,051.00

498,350.90

1,400,000

3,000,000

Hungary

66.33

11.57

83,948.72

15,940.59

52,000

120,000

India

657.90

389.20

200

1586

434,102.60

284,019.30

150,000

1,200,000

Indonesia

166.05

115.33

376

165,641.00

93,216.64

41,000

360,000

Iran

220.67

141.59

533

130,280.40

48,055.64

69,243

238,791

Ireland

31.56

7.89

105,717.90

58,239.50

42,000

230,000

Italy

384.28

45.69

293

464

1,381,282.00

313,361.90

800,000

1,800,000

Japan

1021.26

148.40

759

1242

3,423,077.00

968,846.70

1,700,000

4,800,000

Mexico

270.46

89.39

410

577,179.50

192,926.90

250,000

930,000

Netherlands

164.46

13.80

130

185

457,692.30

133,090.00

270,000

700,000

Norway

30.51

4.97

203,025.60

69,707.47

98,000

320,000

Portugal

38.15

16.38

134,205.10

43,178.44

67,000

200,000

Republic of Korea

262.77

157.11

515

421,666.70

293,045.80

67,000

960,000

Saudi Arabia

176.21

112.26

410

220,852.30

63,307.43

84,439

346,706

South Africa

259.18

54.96

173

388

178,205.10

47,288.13

110,000

290,000

Spain

221.85

62.87

120

344

754,871.80

245,473.50

400,000

1,200,000

Sweden

61.49

12.34

267,179.50

66,488.97

180,000

400,000

Switzerland

41.03

2.38

301,025.60

53,988.05

230,000

410,000

Tunisia

12.36

5.64

18,800.00

9292.90

6300

39,000

Turkey

133.69

65.93

265

285,641.00

128,407.10

120,000

540,000

United Arab Emirates

56.28

41.59

147

98,310.70

53,584.39

24,057

211,638

United Kingdom

550.67

36.24

466

637

1,581,538.00

445,477.50

980,000

2,400,000

United States

5016.18

481.77

4291

5772

8,315,385.00

2,829,593.00

4,400,000

13,000,000

Uruguay

5.05

1.17

13,682.05

3560.15

8800

22,000

Venezuela

105.74

29.28

161

112,461.50

27,658.49

67,000

180,000

Country

Population in hundred thousands

ﬁciency index

Obs.

Mean

Std. dev.

Min

Max

Obs.

Mean

Std. dev.

Min

Max

b
Argentina

324.87

48.71

240

400

105.79

8.29

118

Australia

170.51

25.64

130

220

81.62

13.06

102

Austria

78.18

2.73

80.02

9.37

100

Belgium

100.23

3.48

110

73.79

13.30

103

Brazil

1476.41

293.29

980

1900

74.61

7.39

100

Canada

277.18

35.39

220

340

74.54

14.98

100

Chile

133.00

22.85

170

97.14

11.52

117

China

11,197.44

1537.51

8400

13,000

51.23

32.50

104

Denmark

52.05

1.52

75.13

15.15

101

Finland

49.85

2.15

92.17

11.95

108

France

581.79

35.01

520

650

69.95

14.43

100

Germany

800.26

18.85

780

830

71.72

17.19

100

Hungary

103.59

4.86

100

110

80.59

14.50

100

India

8787.18

1990.00

5700

12,000

72.84

20.18

101

Indonesia

1815.39

358.02

1200

2400

56.18

17.57

100

Iran

523.84

141.47

294

731

199.21

55.17

277

Ireland

36.21

3.81

71.45

15.95

100

Italy

569.23

13.26

540

600

72.48

12.50

102

Japan

1217.95

75.64

1100

1300

81.54

9.41

100

Mexico

838.97

178.23

530

1100

84.77

12.32

100

Netherlands

149.23

10.36

130

170

89.68

12.78

112

Norway

42.90

2.51

88.02

13.89

112

Portugal

99.36

6.26

110

118.09

12.29

135

Republic of Korea

422.82

49.68

330

490

78.10

14.51

103

Saudi Arabia

156.68

62.16

268

430.74

164.49

653

South Africa

356.41

85.76

230

490

85.37

11.20

104

Spain

391.79

29.37

340

460

106.35

7.42

122

Sweden

86.00

3.47

73.00

17.12

100

Switzerland

67.97

4.64

136.90

21.60

167

Tunisia

79.54

16.60

100

93.71

7.56

108

Turkey

539.23

109.24

360

720

98.99

7.57

114

(continued on next page)

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

739

Table 6

reports Pesaran (2007)

[33]

CIPS (cross-sectionally

augmented Im, Pesaran and Shin) unit root test statistics for the
variables in our model for the 36 countries from 1971 to 2009 for
lag orders from 0 to 3. The inclusion of lags enables us to take

potential autocorrelation in the data into account. According to the
results, the variables LN(GDP), LN(EFFICIENCY) and LN(RENEWABLE/
TOTAL ENERGY) are non-stationary while the variables LN(CO

LN(POPULATION) and LN(MANUFACTURING SHARE) are stationary

Table 4 (continued )

Country

Population in hundred thousands

ﬁciency index

Obs.

Mean

Std. dev.

Min

Max

Obs.

Mean

Std. dev.

Min

Max

United Arab Emirates

21.90

16.09

340.65

157.33

552

United Kingdom

576.92

17.64

560

620

71.52

17.39

100

United States

2541.03

304.97

2100

3100

63.01

19.36

100

Uruguay

30.92

1.90

76.28

15.13

103

Venezuela

197.18

52.36

110

280

157.86

24.72

100

200

Country

Share of manufacturing industries in total GDP

Ratio of renewable to total energy

Obs.

Mean

Std. dev.

Min

Max

Obs.

Mean

Std. dev.

Min

Max

c
Argentina

36.96

6.91

5.28

1.09

3.04

7.40

Australia

31.70

5.85

6.78

1.41

5.00

9.80

Austria

32.98

3.37

8.69

2.91

4.07

14.00

Belgium

30.94

5.59

0.65

0.73

0.00

2.82

Brazil

36.13

7.36

32.45

7.30

23.34

53.10

Canada

33.02

2.42

4.96

0.45

4.35

6.38

Chile

38.90

2.60

22.06

3.23

17.65

28.11

China

45.28

1.97

30.24

8.22

14.88

45.38

Denmark

26.29

1.64

3.89

1.93

0.39

7.91

Finland

34.86

2.99

17.75

2.26

14.14

24.92

France

26.98

4.77

6.31

0.62

5.27

7.71

Germany

36.13

5.67

1.66

1.23

0.60

4.91

Hungary

39.09

8.68

3.54

1.05

1.91

5.99

India

25.43

2.07

54.31

9.14

36.63

67.23

Indonesia

39.54

6.25

52.85

12.68

35.55

79.05

Iran

37.01

9.91

0.43

0.14

0.26

0.90

Ireland

35.67

2.58

0.69

0.72

0.00

2.34

Italy

32.47

4.58

1.03

0.57

0.00

3.28

Japan

35.14

4.83

0.73

0.47

0.00

1.29

Mexico

31.63

3.14

9.85

2.66

6.73

17.39

Netherlands

29.43

4.35

0.44

0.00

1.54

Norway

36.48

3.98

4.15

1.81

0.00

6.76

Portugal

28.54

2.43

11.94

2.92

7.94

17.52

Republic of Korea

37.15

4.59

0.18

0.20

0.00

0.71

Saudi Arabia

56.59

11.59

0.02

0.01

0.00

0.05

South Africa

37.71

5.26

15.10

1.17

12.93

17.00

Spain

33.09

4.17

2.62

2.27

0.00

6.44

Sweden

30.59

2.87

12.94

2.11

8.57

17.08

Switzerland

30.15

2.74

3.14

1.52

1.27

5.48

Tunisia

31.74

3.61

16.70

3.55

13.35

26.60

Turkey

28.80

3.64

20.75

8.49

8.28

36.08

United Arab Emirates

52.10

5.68

0.04

0.05

0.00

0.18

United Kingdom

33.09

6.69

0.28

0.34

0.00

1.41

United States

28.06

4.91

3.37

1.07

1.76

5.60

Uruguay

29.51

4.25

22.84

4.46

16.49

32.20

Venezuela

48.75

5.66

1.81

0.63

1.16

3.93

Table 5
Cross-sectional correlation.

Panel A: variables in levels

LN (CO

)

LN (GDP)

LN (POPULATION)

LN (EFFICIENCY)

LN (MANUFACTURING SHARE)

LN (RENEWABLE/TOTAL ENERGY)

avg

0.383

0.952

0.853

0.338

0.215

0.023

avg

0.749

0.952

0.940

0.739

0.536

0.510

60.09

149.18

133.73

53.00

32.47

3.25

p-value

0.000

0.001

Panel B: variables in

ﬁrst difference

LN (CO

)

LN (GDP)

LN (POPULATION)

LN (EFFICIENCY)

LN (MANUFACTURING SHARE)

LN (RENEWABLE/TOTAL ENERGY)

avg

0.156

0.130

0.124

0.086

0.143

0.051

avg

0.213

0.199

0.415

0.173

0.214

0.179

24.12

20.10

19.24

13.37

21.84

6.49

p-value

0.000

Notes: Panels A and B test the variable series in levels and

ﬁrst differences, respectively. avg

and avg

j denote the average and average absolute correlation coefﬁcients across

the N(N

1) sets of correlations. CD is the Pesaran (2004)

[31]

cross-section dependence statistic, which is distributed N(0, 1) under the null of cross-section independence.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

740

when a constant is added in the CADF regression. This result is
robust to the choice of augmentation order. Conversely, when a
constant and trend are added to the CADF regression to capture the
trended nature of the data, the unit root hypothesis cannot be
rejected for alternative choices of lags for all the variables (except
for p

¼ 0 for LN(CO

) and LN(GDP), p

¼ 1 for LN(POPULATION) and

¼ 1 for LN(MANUFACTURING SHARE)). Therefore, taken as a whole,

the results imply that we might have a mixture of I(0) and I(1)
variables. Fortunately, the CCE estimator is robust to lack of coin-
tegration

[11,30]

and

[20]

, which allows us to proceed with using

the variables as they are.

4.2. Econometric results

This section discusses the empirical results. We estimate models

using Pesaran and Smith (1995)'s

[29]

Mean Group (MG) and

Pesaran (2006)'s

[32]

CCE (Common Correlated Effects) estimators

(

Table 7

). These two estimators differ in their assumptions on cross-

sectional dependence: the former does not allow for cross-
sectional dependence while the latter does. We also perform a
number of diagnostic tests including residual cross-correlation
tests and residual stationarity tests and display the RMSE (root
mean squared error) statistic as a measure of goodness of

ﬁt.

As shown by the diagnostic tests on residual cross-correlation,

the MG estimator suffers cross-sectionally dependent residuals:
according to the diagnostic test results, cross-sectional indepen-
dence is rejected at the 1% signi

ﬁcance level. This result is also

consistent with the

ﬁndings of the preliminary spatial dependence

tests. Thus, taken as a whole, the possibility of cross-sectional
dependence is an important empirical issue. In line with this
observation, once the results of the diagnostic tests on the CCE
estimator residuals are checked, it is seen that there is no evidence
of cross-section dependence, providing support for our choice of
the CCE estimator. Furthermore, in the CCE estimation (as well as in
the MG estimation) the null of residual non-stationarity is rejected,
providing evidence of residual stationarity, which is crucial for a
good

ﬁtting econometric model. Finally, the estimation with the

CCE estimator results in a lower root mean square error (26%
smaller than what the MG estimator produces).

Before interpreting the country-speci

ﬁc results obtained from

the CCE, we examine the long-run average coef

ﬁcients. The esti-

mated income coef

ﬁcient is 0.6219 and 0.4178 in the MG estimator

model and CCE model, respectively, and in both models the coef-
ﬁcient is statistically signiﬁcant at the 1% level. In the ﬁrst column
of

Table 7

, the magnitude of the statistically signi

ﬁcant (at the 1%

level) coef

ﬁcient of efﬁciency index (LN (EFFICIENCY)) is 0.6724.

When controlling for cross-sectional dependence, the coef

ﬁcient of

ﬁciency is moderately lower (0.5520 in the second column). The

coef

ﬁcient on the share of manufacturing industries of total GDP

(LN(MANUFACTURING SHARE)) is positive (0.14 and 0.12) and sta-
tistically signi

ﬁcant at the 5% level in both speciﬁcations. Further-

more, the renewable energy consumption has quite a large
in

ﬂuence on CO

emissions: the coef

ﬁcient of LN(RENEWABLE/TO-

TAL ENERGY) is negative and signi

ﬁcant (at the 1% level for the MG

estimator and at the 5% level for the CCE estimator), ranging
between

0.13 and 0.11 across two estimators. Finally, the pop-

ulation variable (LN(POPULATION)) makes statistically insigni

ﬁcant

contributions to the regression equations.

As our model is a log

elog model, the coefﬁcients obtained could

be interpreted as elasticities (

Table 7

). A 1% increase in GDP is

associated with a 0.42% increase in CO

emissions. In other words, a

doubling of GDP brings about a 42% increase in total CO

emissions.

Another factor contributing to CO

emission increases is the share

of manufacturing industries: a 1% increase in the share of
manufacturing industries increases CO

emissions by 0.12%,

everything else equal. On the other hand, energy ef

ﬁciency and the

share of renewables in total energy have quite strong implications
for CO

emissions abatement: a 1% increase in energy ef

ﬁciency

reduces CO

emissions by 0.55% in the long-run, ceteris paribus.

Elsewhere, a 1% increase in the ratio of renewable to total energy
results in a 0.11% reduction in CO

emissions.

Table 8

reports the estimated results for each of the 36 coun-

tries. There is a large degree of heterogeneity across countries in
this sample. To begin with the income coef

ﬁcient, it is signiﬁcant for

emerging economies at 1% level such as Argentina, Brazil, Chile,
Iran, Mexico, Korea, and at 5% level for Tunisia, and Venezuela. In
contrast, for developed countries the income coef

ﬁcient is mostly

insigni

ﬁcant except for Australia, Canada, Italy, Norway, Portugal,

and United States (with 1% signi

ﬁcance level). The long-term rela-

tionship between income and CO

emissions is the strongest for

Chile and Republic of Korea: A one % increase in gross domestic
product is linked to an increase of approximately 1.2% and 1.3% in
total CO

emissions, respectively. In general, all of the countries

with signi

ﬁcant income and CO

relationship have positive

Table 6
Pesaran (2007)

[33]

CIPS test of unit root.

Panel A: variables in levels

Lags

LN (CO

)

LN (GDP)

LN (POPULATION)

LN (EFFICIENCY)

LN (MANUFACTURING SHARE)

LN (RENEWABLE/TOTAL ENERGY)

Constant
0

4.904 (0.00)

3.083 (0.00)

0.196 (0.42)

2.122 (0.98)

2.301 (0.01)

1.451 (0.93)

2.741 (0.00)

0.986 (0.16)

11.061 (0.00)

1.103 (0.87)

3.103 (0.00)

0.847 (0.80)

1.939 (0.03)

0.562 (0.29)

4.568 (0.00)

0.868 (0.81)

2.105 (0.02)

0.546 (0.29)

1.911 (0.03)

0.577 (0.28)

7.787 (0.00)

0.854 (0.80)

2.916 (0.00)

0.045 (0.48)

Constant and trend
0

3.168 (0.00)

1.657 (0.05)

6.313 (1.00)

0.167 (0.57)

0.012 (0.51)

0.100 (0.54)

0.841 (0.20)

0.836 (0.80)

6.774 (0.00)

1.020 (0.85)

1.832 (0.03)

0.738 (0.77)

0.015 (0.49)

1.355 (0.91)

1.859 (0.97)

0.573 (0.72)

0.060 (0.52)

0.622 (0.73)

0.016 (0.50)

1.657 (0.95)

0.945 (0.17)

1.445 (0.93)

1.469 (0.07)

0.249 (0.60)

Panel B: variables in

ﬁrst difference

Lags

LN (CO

)

LN (GDP)

LN (POPULATION)

LN (EFFICIENCY)

LN (MANUFACTURING SHARE)

LN (RENEWABLE/TOTAL ENERGY)

24.770 (0.00)

23.496 (0.00)

0.789 (0.79)

23.122 (0.00)

19.194 (0.00)

21.866 (0.00)

14.335 (0.00)

14.167 (0.00)

6.896 (0.00)

13.958 (0.00)

14.722 (0.00)

10.898 (0.00)

9.364 (0.00)

8.906 (0.00)

1.550 (0.06)

9.516 (0.00)

7.710 (0.00)

5.175 (0.00)

7.508 (0.00)

6.384 (0.00)

2.350 (0.01)

7.877 (0.00)

5.934 (0.00)

3.518 (0.00)

Notes: We report the standardized Z-tbar statistic and its p-value (in parenthesis). The null hypothesis is that all series are non-stationary. Lags show the lag augmentation in
the Dickey Fuller regression. In Panel A, we augment the Dickey Fuller regression for variables in levels with a constant or a constant and trend; while in Panel B we only
employ a drift (constant) for the variables in

ﬁrst differences.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

741

elasticities meaning that as their income increase the CO

emis-

sions increase after taking into account changes in economic
structure, energy ef

ﬁciency and renewable energy use. The

responsiveness of CO

emissions to the changes in income depends

on level of demand and supply of environmental quality in each
country. There might be more room in some countries than others
in the demand and supply of CO

reduction.

In addition to being mostly positive, the ef

ﬁciency coefﬁcient is

statistically signi

ﬁcant at the 5% for 7 countries and at the 1% for 17

countries. Among the statistically signi

ﬁcant effects, the most

pronounced effect of ef

ﬁciency increases on CO

emissions is seen

for Chile, followed by Denmark. A one percent increase in ef

ﬁciency

is associated with CO

reductions of 1.1% and 1% for Chile and

Denmark, respectively. Since 2005 Chile has been implementing
various measures to increase energy ef

ﬁciency. For private com-

panies there are three instruments, pre-investment in energy ef

ﬁ-

ciency, preferential interest rate and a guarantee system for energy
ef

ﬁciency projects. National Energy Efﬁciency Program funding

grew from $1 million USD to $34 million USD, and facilitated other
measures over the period of 2006

e2009. The Country has strong

energy labeling program and Energy Ef

ﬁciency Agency has engaged

in activities for commercial, public, residential, and industry and
mining sectors

[24]

. Denmark has passive energy house and zero

energy buildings. With a revised strategy all new buildings must

have met the standard of low energy houses by 2010 which will
reduce the energy consumption of new houses by 60% compared to
actual standards which targets to reach plus energy houses in the
long-run

[28]

Changes in the share of manufacturing industries in total GDP

do not seem to affect CO

emissions for the great majority of

countries in our sample. Shifts to or from manufacturing industries
bring about statistically signi

ﬁcant changes at the 1% level and at

the 5% level in CO

emissions for only 2 and 8 countries, respec-

tively. Furthermore, among the statistically signi

ﬁcant relations, the

magnitude is the largest for Canada and Sweden: a one percent
increase in the share of manufacturing industries is related to in-
creases of 0.49% and 0.44% of CO

emissions for Canada and Swe-

den, respectively. Manufacturing tend to be more pollution-
intensive than either agriculture or services. There might be
certain range of the share of industrial sector where the emissions
might be higher than that of other ranges

[27]

. There are also

several factors such as capital and labor intensities of industry
sectors, trade relationships between countries, differences in
stringencies of environmental regulations across countries, foreign
direct investment affect the direction of the impact of change in
economic structure on CO

emissions. Demand for goods produced

in different sectors and differentiated rates of technological prog-
ress on a sectoral level are the driving forces behind the patterns of
sectoral change

[12]

. Therefore, the differences between the

countries need to be investigated via country level studies.

Finally, changes in the renewable energy consumption have

effects on CO

emissions with varying magnitudes for 15 out of 36

countries. The most pronounced effect of adopting renewable en-
ergy is for India and China: a one percent increase in the ratio of
renewable energy use to total energy use is associated with 1.7%
and 1.3% decreases of CO

emissions reductions for India and China,

respectively. For China, speci

ﬁc discussions and recommendations

are clari

ﬁcation of the feed-in tariff system, greater transparency in

the Renewable Energy Fund System, adoption of smart grid tech-
nology, increase support for renewable energy and R

&D and

demonstration activities

[36]

. For India, to enhance the deployment

of renewable energy the recommendations focus on i. improving
the regulatory set up for RPO (renewable purchase obligations),
RECs (renewable energy certi

ﬁcates), and ensuring policy certainty,

ii.

ﬁnancial and budgetary issues including introduction of green

bonds, enhancing the utilization of National Clean Energy Fund,
increasing the budget of the Ministry of New and Renewable En-
ergy, substituting kerosene with renewable energy, and using
corporate social responsibility funds for renewable energy, and iii.
ﬁscal provisions via tax incentives for distributed renewable energy
generation and improve direct and indirect tax bene

ﬁts to the

targets

[17]

To take an advantage of the signi

ﬁcant power of

renewable energy in reducing CO

emissions, country speci

ﬁc

studies are needed. Illustratively, for the road transportation sector
in China, He and Chen (2013)

[15]

suggest a policy option of

adjustment of renewable portfolio standards to account for po-
tential off-peak charging of electric vehicles. They infer that with a
low-carbon electricity system, plug-in hybrid vehicles could sub-
stantially reduce GHGs as well as oil dependence.

5. Conclusion and policy implications

The only way to slow down global warming, one of the biggest

challenges facing the worldwide population, is to reduce GHG
emissions. Thus, there is increasing pressure to develop successful
policies for the control and prevention of CO

emissions. Energy

ﬁciency and renewable energy are the two indispensable solu-

tions for emission reduction and prevention. Even though the
importance of these two solutions are emphasized by several

Table 7
Heterogeneous panel estimation results for the model.

Dep. var.: LN(CO

)

Mean group
estimator

Common correlated
effects estimator

LN(GDP)

0.6219*** (0.080)

0.4178*** (0.073)

LN(POPULATION)

0.3288 (0.502)

0.6941 (0.431)

LN(EFFICIENCY)

0.6724*** (0.074)

0.5520*** (0.054)

LN(MANUFACTURING SHARE)

0.1357** (0.060)

0.1161** (0.046)

LN(RENEWABLE/TOTAL ENERGY)

0.1305*** (0.037)

0.1051** (0.043)

CONSTANT

21.5476*** (7.774) 20.5855*** (7.462)

Trends

Included

Wald Statistic and p-value for

Joint signi

ﬁcance test

160.74 (0.000)

151.11 (0.000)

Root mean squared error

0.0343

0.0253

CD test statistic and p-value for

cross-sectional dependence
of residuals test

4.52 (0.000)

0.64 (0.521)

IPS test on residual stationarity

ADF(0)

16.705***

27.033***

ADF(1)

11.672***

18.570***

ADF(2)

7.803***

13.667***

ADF(3)

6.760***

10.656***

CIPS test on residual stationarity

ADF(0)

13.675***

21.019***

ADF(1)

8.790***

14.766***

ADF(2)

5.428***

10.228***

ADF(3)

4.258***

6.982***

Observations

1237

Countries

Notes:

Robust standard errors associated with t-statistics for Mean Group and Common

Correlated effects estimates are in brackets. The estimates are weighted means. ***,
** and * indicate the 1%, 5% and 10% signi

ﬁcance levels, respectively.

p-values associated with

2 test statistics for the joint signi

ﬁcance of the vari-

ables are in brackets.

p-values associated with Pesaran (2004)

[31]

CD test statistic for the null of

cross-sectionally independent residuals are in brackets.

IPS denotes Im-Pesaran-Shin. The test statistics for the null of nonstationarity of

the residuals and the corresponding signi

ﬁcance levels are displayed. Numbers in

brackets denote the lags used in the augmented regression. ***, ** and * indicate the
1%, 5% and 10% signi

ﬁcance levels, respectively.

The standardized Z-tbar statistic for the null of nonstationarity of the residuals

and the corresponding signi

ﬁcance levels are displayed. Numbers in brackets denote

the lags used in the augmented regression. We augment the Dickey Fuller regression
for the residuals with a constant and trend. ***, ** and * indicate the 1%, 5% and 10%
signi

ﬁcance levels, respectively.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

742

studies

[19,38]

, their role in curbing CO

emissions are hardly

investigated in the literature. The paper develops an energy ef

ﬁ-

ciency measure for each country in the sample and empirically
investigates the roles that energy ef

ﬁciency, renewable energy and

industrialization play in reducing CO

emissions for a group of 36

developed and developing countries over the period 1971

e2009.

The CCE estimation model is used to take cross-section dependence
into account and to allow for heterogeneity by estimating a
different parameter vector for each country in the sample.

The empirical

ﬁndings in the paper provide evidence for a long-

term relationship among energy ef

ﬁciency, renewable energy,

industrialization, income, and CO

emissions. The IEA (Interna-

tional Energy Agency) (2010)

[18]

emphasizes that

‘Increasing en-

ergy ef

ﬁciency, much of which can be achieved through low-cost

options, offers the greatest potential for reducing CO

emissions over

the period to 2050. It should be the highest priority in the short term.

’

Our empirical

ﬁndings support the expected effect and the

importance of energy ef

ﬁciency on CO

emissions. A one percent

increase in energy ef

ﬁciency reduces CO

emissions by 0.55% over

the long term. For a large number of countries in our sample, 24 out
of 36, policies related to energy ef

ﬁciency seem to be effective in

Table 8
Common correlated effects estimate of the coef

ﬁcients by country.

Country

LN(GDP)

LN(POPULATION)

Coef

ﬁcient

Std. err.

Coef

ﬁcient

Std. err.

a
Argentina

0.5913***

0.098

0.9978

1.120

Australia

0.5895***

0.214

0.5455

0.825

Austria

0.1233

0.277

1.8679**

0.826

Belgium

0.4338

0.498

6.5680

4.359

Brazil

1.0298***

0.132

1.8538***

0.435

Canada

0.4339***

0.134

0.2632

1.100

Chile

1.2644***

0.256

1.8469

2.061

China

0.0607

0.096

1.1237***

0.264

Denmark

0.3268

0.485

1.1323

2.155

Finland

0.3268

0.309

13.4711**

5.669

France

0.2005

0.270

3.8717

3.115

Germany

0.0745

0.194

0.8656

0.869

Hungary

0.6010***

0.133

4.1550***

1.557

India

0.1374

0.235

3.3699***

0.967

Indonesia

0.2419

0.179

2.2434***

0.810

Iran

0.8586***

0.068

0.2721

0.388

Ireland

0.7229***

0.280

2.6135

1.873

Italy

0.4391***

0.144

1.3697**

0.641

Japan

0.2909

0.280

2.1022

2.111

Mexico

0.6470***

0.111

1.4552***

0.531

Netherlands

0.0565

0.240

2.1582

3.670

Norway

1.0031***

0.206

5.6302***

1.754

Portugal

1.1165***

0.172

1.3185*

0.778

Republic of Korea

1.1250***

0.284

0.2438

2.920

Saudi Arabia

0.5557***

0.143

1.5281***

0.377

South Africa

0.1986

0.216

0.3339

0.625

Spain

0.2377

0.415

0.7011

0.898

Sweden

0.1535

0.300

1.9721

1.607

Switzerland

0.4395

0.443

0.4525

2.526

Tunisia

0.6113**

0.251

1.0542

0.701

Turkey

0.3607

0.240

6.4136***

1.415

United Arab Emirates

0.2815

0.307

0.7005**

0.328

United Kingdom

0.3750

0.359

7.0794

7.351

United States

0.6822***

0.130

0.4049

0.919

Uruguay

0.5184

0.685

27.2221***

9.373

Venezuela

0.4144**

0.165

0.5092

0.880

Country

LN(EFFICIENCY)

LN(MANUFACTURING SHARE)

Coef

ﬁcient

Std. err.

Coef

ﬁcient

Std. err.

b
Argentina

0.2513

0.168

0.0071

0.091

Australia

0.5562**

0.219

0.1223*

0.071

Austria

0.9542***

0.156

0.0157

0.375

Belgium

0.7263**

0.360

0.1699

0.327

Brazil

0.7208***

0.226

0.0942*

0.049

Canada

0.2367

0.157

0.4852***

0.109

Chile

1.1365***

0.384

0.2559

0.166

China

0.1384**

0.059

0.1107

0.079

Denmark

0.9917***

0.307

0.8217*

0.458

Finland

0.3895

0.352

0.4602

0.385

France

0.7150**

0.351

0.2866

0.498

Germany

0.3750***

0.121

0.4141**

0.195

Hungary

0.1994*

0.107

0.0649

0.108

India

0.0156

0.297

0.1740

0.143

Indonesia

0.2578

0.219

0.1723

0.140

Iran

0.7358***

0.128

0.0967**

0.044

Ireland

0.7620***

0.293

0.0711

0.313

Italy

0.4445***

0.162

0.3915

0.356

Japan

0.4945*

0.254

0.1801

0.348

Mexico

0.3279***

0.126

0.0045

0.059

Netherlands

0.5556***

0.197

0.2945

0.314

Norway

0.1359

0.285

0.2569**

0.112

Portugal

0.9790***

0.278

0.0949

0.221

Republic of Korea

0.8719*

0.483

0.2372

0.367

Saudi Arabia

0.7256***

0.053

0.0923

0.115

South Africa

0.7032***

0.171

0.6356***

0.168

Spain

0.8895**

0.434

0.0412

0.525

Sweden

0.6586***

0.168

0.4381**

0.186

Switzerland

0.6243**

0.274

0.5217

0.455

Tunisia

0.0400

0.333

0.3654**

0.148

Turkey

0.3799*

0.206

0.2946**

0.137

Table 8 (continued )

Country

LN(EFFICIENCY)

LN(MANUFACTURING SHARE)

Coef

ﬁcient

Std. err.

Coef

ﬁcient

Std. err.

United Arab Emirates

0.4598***

0.102

0.2323*

0.138

United Kingdom

0.7644**

0.307

0.0635

0.433

United States

0.5625***

0.187

0.0988

0.124

Uruguay

0.7296

1.037

0.2260

0.601

Venezuela

0.5115***

0.194

0.0532

0.092

Country

LN(RENEWABLE/TOTAL ENERGY)

Coef

ﬁcient

Std. err.

c
Argentina

0.1154**

0.051

Australia

0.0114

0.032

Austria

0.1137***

0.037

Belgium

0.0244

0.050

Brazil

0.8138***

0.101

Canada

0.0102

0.062

Chile

0.2965

0.227

China

1.2640***

0.077

Denmark

0.0089

0.067

Finland

0.4280**

0.201

France

0.4090***

0.115

Germany

0.0027

0.034

Hungary

0.0040

0.041

India

1.7441***

0.323

Indonesia

0.6629**

0.286

Iran

0.0174

0.030

Ireland

0.0652

0.122

Italy

0.0827***

0.030

Japan

0.1373*

0.079

Mexico

0.1848**

0.076

Netherlands

0.0320

0.052

Norway

0.0774

0.097

Portugal

0.0801

0.055

Republic of Korea

0.0045

0.028

Saudi Arabia

0.0317

0.024

South Africa

0.4561**

0.197

Spain

0.0380*

0.020

Sweden

0.5741***

0.081

Switzerland

0.0251

0.038

Tunisia

0.4948*

0.297

Turkey

0.4512**

0.224

United Arab Emirates

0.1437*

0.081

United Kingdom

0.0917*

0.056

United States

0.0118

0.020

Uruguay

0.9687**

0.438

Venezuela

0.2890***

0.079

Notes: ***, ** and * indicate the 1%, 5% and 10% signi

ﬁcance levels, respectively.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

743

curbing CO

emissions. The greatest impact of energy ef

ﬁciency on

emission reduction is observed in the Chile, followed by Denmark.
Developing countries might bene

ﬁt from preferential interest rates

and guarantee system for energy ef

ﬁciency companies, and

implement strong labeling program with speci

ﬁc institutional set

up such as energy ef

ﬁciency agency. Developed countries might

review the performance of their residential sector once suf

ﬁcient

improvements in other sectors such as agriculture, industry and
mining has been reached. Speci

ﬁcally, setting strategies to trans-

form to plus energy house in the long-run, might be plausible ones,
as they are for Germany and Denmark.

Renewable energy is promoted throughout the world for several

reasons. Renewable policies are mainly driven by environmental
concerns, and they aim to reduce CO

emissions and local pollut-

ants. Renewable energy is also attractive due to the resulting eco-
nomic

stimulation,

enhancement

energy

security

and

diversi

ﬁcation of the energy supply. In 2010, renewable energy

accounted for 13% of global primary energy demand. In the power
sector, the number of countries implementing renewable energy
technologies is expected to exceed 70 by 2017. IEA projects an in-
crease in the share of renewable energy in primary energy use,
mainly due to governmental support, CO

pricing in certain regions,

rising fossil fuel prices in the long term and falling costs for
renewable technologies. Under the same projections, CO

emis-

sions are expected to decrease by over 4.1 Gt by 2035 (IEA, 2012).
Unlike Apergis et al. (2010)

[5]

; we

ﬁnd that renewable energy

plays a signi

ﬁcant role in reducing CO

emissions; a 1 percent in-

crease in the use of renewable energy in total energy is related to an
average decrease of 0.11 percent of CO

emissions over the long

term. We observe signi

ﬁcant effects of renewable energy in 15

countries out of 36, with the India and China exhibiting the largest
effect. As stated in Chiu and Chang (2009)

[10]

; a threshold for

effectiveness may exist for some countries. As implemented in
some developed countries, other nations can also various means to
support the use of renewable energy, including renewable portfolio
standards for electricity production; renewable fuel standards for
transportation fuels; production tax credits for wind, solar, bio-
diesel and biofuels; and tax incentives, grants and loans. As an
example to other countries, China has introduced Chinese pro-
grams for building-integrated PV installation, large-scale solar PV
power stations, rural utilization of solar water heaters and the
construction of gigawatt-level wind farm bases can be imple-
mented in other countries

[36]

. Adjusting renewable portfolio

standards to account for potential off-peak charging of electric
vehicles to reduce GHGs and oil dependence is a plausible policy
alternative in transportation industry

[15]

. Furthermore, attaining

policy certainty, improving regulatory framework for renewable
purchase obligations, introducing renewable energy certi

ﬁcates,

using national funds effectively, using borrowing mechanisms such
as green bonds, enabling appropriate tax incentives for distributed
renewable energy generation are some other recommended policy
measures.

Economies move among the agricultural, industrial and service

sectors and grow in output over different time spans. While the
relationship between economic activities and CO

emissions is

already complex in a static world, this relationship becomes even
more complex in a dynamic world. Consistent with the literature,
we

ﬁnd that an increase in income increases CO

emissions for 17

countries in our sample, with the largest effect observed for Chile
followed by Republic of Korea. We also do

ﬁnd evidence of the

impact of industrialization on emissions for the sample, although
for only 5 countries we observe signi

ﬁcant positive effects, with the

largest effect being observed for Canada and Sweden. On the other
hand, for 3 countries in our sample, we

ﬁnd a negative and sig-

ﬁcant effect of changes in the composition of economic activities

on CO

emissions over the long-run, which shows that the sign of

these effects varies from country to country. A negative relationship
may indicate the adoption of clean industrial production processes
or the acceptance of the pollution heaven hypothesis, while a
positive relationship may be related to the use of domestic dirty
industrial processes. It may also be a result of an increase in the
share of industrial sector that are labor intensive and relatively less
pollution intensive than capital intensive sectors. Differences in
stringencies of environmental regulations and resources allocated
to enforcement of these regulations, trade and foreign direct in-
vestment patters, domestic demand for the dirty and clean indus-
trial products, and patterns of technological change in each
subsector will also play a role in shaping the effect of industriali-
zation on CO

emissions. Moreover, several factors such as demand

for goods produced in different sectors and differentiated rates of
technological progress on a sectoral level are the driving forces
behind the patterns of sectoral change (Groot, 1998). Policy rec-
ommendations under the complex relationships amongst afore-
mentioned factors require country and sector speci

ﬁc studies.

Illustratively, He and Chen (2013)

[15]

study China's road trans-

portation sector to address possible policy measures for the
reduction of energy consumption and GHG emissions. Chen and He
(2013)

[8]

study the impacts of deregulation of the electricity

generation sector and retailing activities on the macroeconomy and
electricity users. Therefore, potential future studies could be on the
role of energy ef

ﬁciency and renewable energy in a speciﬁc coun-

try/sector.

Acknowledgments

A preliminary version of this work was presented at the Man-

nheim Energy Conference 2013, Sustainable Energy Markets, ZEW
(Center for European Economics Research), and MaCCI (Mannheim
Center for Competition and Innovation), Mannheim, Germany, June
24

e25, 2013 and at the IEC-TEA 2014 (International Conference on

Economics-Turkish

Economic

Association),

Antalya,

Turkey,

October 18-20, 2014. We appreciate the comments of the confer-
ence participants. Any errors or omissions are the responsibility of
the authors.

References

[1] Al-mulali U, Binti Che Sab CH. The impact of energy consumption and CO

emission on the economic growth and

ﬁnancial development of in the Sub-

Saharan African countries. Energy 2012;39:180

e6.

http://dx.doi.org/10.1016/

j.energy.2012.01.032

[2] Al-mulali U. Factors affecting CO

emission in the Middle East: a panel data

analysis. Energy 2012;44:564

e9.

http://dx.doi.org/10.1016/j.energy.2012.05.

045

[3] Al-mulali U. Oil consumption, CO

emission and economic growth in MENA

countries. Energy 2011;36:6165

e71.

http://dx.doi.org/10.1016/j.energy.2011.

07.048

[4] Andrews DWK. Cross-section regression with common shocks. Econometrica

2005;73:1551

e85.

http://dx.doi.org/10.1111/j.1468-0262.2005.00629.x

[5] Apergis N, Payne JE, Menyah K, Wolde-Rufael Y. On the casual dynamics

between emissions, nuclear energy, renewable energy, and economic
growth. Ecol Econ 2010;69:2255

e60.

http://dx.doi.org/10.1016/j.enpol.2009.

09.002

[6] Arouri MEH, Youssef AB, M'henni H, Rault C. Energy consumption, economic

growth and CO

emissions in Middle East and North African countries. Energy

Policy 2012;45:342

e9.

http://dx.doi.org/10.1016/j.enpol.2012.02.042

[7] Cavalcanti TVV, Mohaddes K, Raissi M. Growth, development and natural

resources: new evidence using a heterogeneous panel analysis. Q Rev Econ
Finance 2011;51:305

e18.

http://dx.doi.org/10.1016/j.qref.2011.07.007

[8]

Chen S-M, He L-Y. Deregulation or governmental regulation? A counterfactual
perspective on China's electricity market reform. China World Econ
2013;20(4):101

e20

[9] Cherniwchan J. Economic growth, industrialization and the environment.

Resour Energy Econ 2012;34:442

e67.

http://dx.doi.org/10.1016/j.reseneeco.

2012.04.004

[10] Chiu CL, Chang TH. What proportion of renewable energy supplies is

needed to initially mitigate CO

emissions in OECD member countries.

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

744

Renew Sustain Energy Rev 2009;13:1669

e74.

http://dx.doi.org/10.1016/j.

rser.2008.09.026

[11] Chudik A, Pesaran MH, Tosetti E. Weak and strong cross section dependence

and estimation of large panels. Econ J 2011;14:C45

e90.

http://dx.doi.org/10.

1111/j.1368-423X.2010.00330.x

[12]

De Groot HLF. Structural change, economics growth and the environmental
kuznets curve a theoretical perspective. In: OCFEB research memorandum
9911,

‘Environmental policy, economic reform and endogenous technology’;

2003. Working Paper Series 1

[13] Dinda S. Environmental kuznets curve hypothesis: a survey. Ecol Econ

2004;49:431

e55.

http://dx.doi.org/10.1016/j.ecolecon.2004.02.011

[14] Esseghir A, Khouni LH. Economic growth, energy consumption and sus-

tainable development: the case of the Union for the Mediterranean coun-
tries. Energy 2014:218

e25.

http://dx.doi.org/10.1016/j.energy.2014.04.050

[15]

He L-H, Chen Y. Thou shalt drive electric and hybrid vehicles: scenario anal-
ysis on energy saving and emission mitigation for the road transportation
sector in China. Transp Policy 2013;25:30

e40

[16] Holly S, Pesaran MH, Yamagata T. A spatio-temporal model of house prices in

the USA. J Econ 2010;158:160

e73.

http://dx.doi.org/10.1016/j.jeconom.2010.

03.040

[17]

Idam Infrstructure Advisory Private Limited. Re-energising India: policy,
regulatory and

ﬁnancial initiatives to augment renewable energy deployment

in India. Climate Parliement and Idam; 2013

[18] International Energy Agency. Energy Technologies Perspectives. Scenarios and

strategies to 2010. Executive summary 2010. 2010.

http://www.iea.org/

techno/etp/etp10/English.pdf

[accessed 04.04.13].

[19] Jaunky VC. The CO

emissions-income nexus: evidence from rich countries.

Energy Policy 2011;39:1228

e40.

http://dx.doi.org/10.1016/j.enpol.2010.11.

050

[20] Kapetanios G, Pesaran MH, Yamagata T. Panels with nonstationary multifactor

error

structures.

Econ

2011;160:326

e48.

http://dx.doi.org/10.1016/j.

jeconom.2010.10.001

[21] Lee CC, Chang CP. Energy consumption and economic growth in Asian

economies: a more comprehensive analysis using panel data. Resour Energy
Econ 2008;30:50

e65.

http://dx.doi.org/10.1016/j.reseneeco.2007.03.003

[22] Metcalf GE. An empirical analysis of energy intensity and its determinants at

the state level. Energy J 2008;29:1

e26.

http://dx.doi.org/10.5547/ISSN0195-

6574-EJ-Vol29-No3-1

[23] Narayan PK, Smyth R. Multivariate Granger causality between electricity

consumption, exports and GDP: evidence from a panel of Middle Eastern
countries. Energy Policy 2009;37:229

e36.

http://dx.doi.org/10.1016/j.enpol.

2008.08.020

[24]

Morgan P, Martinez D, Maxwell A. From good to great: the next step in chilean
energy ef

ﬁciency. Natural Resources Defense Council (NDRC); March 2014

[25]

Nillesen PHL, Haffner RCG, Ozbugday FC. A global perspective on the long-
term impact of increased energy ef

ﬁciency. In: Sioshansi FP, editor. End of

electricity demand growth: how energy ef

ﬁciently can put an end to the need

for more power plants. Elsevier; 2013

[26] Ozturk I. A literature survey on energy-growth nexus. Energy Policy 2010;38:

340

e9.

http://dx.doi.org/10.1016/j.enpol.2009.09.024

[27]

Panayotou T. Demystifying the environmental kuznets curve: turning a black
box into a policy tool. Environ Dev Econ 1997;2(04):465

e84

[28]

Pasquier SB. Implementation of the 25 energy ef

ﬁciency policy recommen-

dations in IEA member countries in IEA member countries: recent de-
velopments. Energy ef

ﬁciency series, IEA/OECD series. 2011

[29] Pesaran MH, Smith RP. Estimating long-run relationships from dynamic het-

erogeneous panels. J Econ 1995;68:79

e113.

http://dx.doi.org/10.1016/0304-

4076(94http://dx.doi.org/10.1016/0304-4076(94)01644-F

[30] Pesaran MH, Tosetti E. Large panels with common factors and spatial corre-

lation. J Econ 2011;161:182

e202.

http://dx.doi.org/10.1016/j.jeconom.2010.

12.003

[31]

Pesaran MH. General diagnostic tests for cross section dependence in panels.
Cambridge Working Papers in Economics, No. 0435. University of Cambridge;
June 2004

[32] Pesaran MH. Estimation and inference in large heterogeneous panels with a

multifactor error structure. Econometrica 2006;74:967

e1012.

http://dx.doi.

org/10.1111/j.1468-0262.2006.00692.x

[33] Pesaran MH. A simple panel unit root test in the presence of cross section

dependence. J Appl Econ 2007;22:265

e312.

http://dx.doi.org/10.1002/jae.951

[34] Piaggio M, Padilla E. CO

emissions and economic activity: heterogeneity

across countries and non-stationary series. Energy Policy 2012;46:370

e81.

http://dx.doi.org/10.1016/j.enpol.2012.03.074

[35] Poa H, Yu H, Yang Y. Modeling the CO

emissions, energy use, and economic

growth in Russia. Energy 2011;36:5094

e100.

http://dx.doi.org/10.1016/j.

energy.2011.06.004

[36]

REN21. Recommendations for improving the effectiveness of renewable en-
ergy policies in China, renewable energy policy network for the 21st century.
October 2009

[37] Saboori B, Sulaiman J. CO

emissions, energy consumption and economic

growth in Association of Southeast Asian Nations (ASEAN) countries: a coin-
tegration approach. Energy 2013;55:813

e22.

http://dx.doi.org/10.1016/j.

energy.2013.04.038

[38] Sadorsky P. The effect of urbanization on CO

emissions in emerging econo-

mies. Energy Econ 2014;41:147

e53.

http://dx.doi.org/10.1016/j.eneco.2013.

11.007

[39] Wagner M. The carbon kuznets curve: a cloudy picture emitted by bad

econometrics? Resour Energy Econ 2008;30:388

e408.

http://dx.doi.org/10.

1016/j.reseneeco.2007.11.001

[40] Wang YC. Functional sensitivity of testing the environmental kuznets curve

hypothesis. Resour Energy Econ 2013;35:451

e66.

http://dx.doi.org/10.1016/j.

reseneeco.2013.01.003

[41] Wing IS. Explaining the declining energy intensity of the US economy. Resour

Energy Econ 2008;30:21

e49.

http://dx.doi.org/10.1016/j.reseneeco.2007.03.

001

F.C. €

Ozbu

gday, B.C. Erbas / Energy 82 (2015) 734e745

745

Document Outline

How effective are energy efficiency and renewable energy in curbing CO2 emissions in the long run? A heterogeneous panel da ...

Wyszukiwarka

Podobne podstrony:
Tea consumption and cardiovascular risk in the SU VI MAX study Are life style factors important
There are a lot of popular culture references in the show
Greenshit go home Greenpeace, Greenland and green colonialism in the Arctic
Civil Society and Political Theory in the Work of Luhmann
54 767 780 Numerical Models and Their Validity in the Prediction of Heat Checking in Die
No Man's land Gender bias and social constructivism in the diagnosis of borderline personality disor
Luhmann's Progeny Systems Theory and Literary Studies in the Post Wall Era
The Immigration Experience and Converging Cultures in the U
[13]Role of oxidative stress and protein oxidation in the aging process
Feminism And Gender Equality In The90's
Keynesian?onomic Theory and its?monstration in the New D
Witch,fairy and folktale narratives in the trial of Bessie Dunlop
Warneck; Descent of Socrates Self knowledge And Cryptic Nature in the Platonic Dialogues
Unsolved Mysteries An Exhibition of Unsolved Mysteries and Enigmatic Findings in the History of Hum
Confirmation of volatiles by SPME and GC MS in the investiga
Wright Material and Memory Photography in the Western Solomon Islands
Co existence of GM and non GM arable crops the non GM and organic context in the EU1

więcej podobnych podstron

How effective are energy efficiency and renewable energy in curbing CO2 emissions in the long run A heterogeneous panel data analysis

Document Outline