Energy Policy 105 (2017) 386–397

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

Energy innovation and renewable energy consumption in the correction of MARK

air pollution levels
⁎
Agustin Alvarez-Herranza, Daniel Balsalobre-Lorenteb, , Muhammad Shahbazc, José
María Cantosb
a
University of Castilla-La Mancha, Dept. of Spanish and International Economics, Econometrics and Economic History and Institutions, Spain
b
University of Castilla-La Mancha, Dept. of Political Economy and Public Finance, Economic and Business Statistics and Economic Policy, Spain
c
Energy and Sustainable Development, Montpellier Business School, Montpellier, France

A R T I C L E I N F O A BS T RAC T

JEL classification: This study analyses the relationship between economic growth and environmental pollution. Specifically, it
Q41 investigates the presence of an environmental Kuznets curve (EKC) in 17 OECD countries over the period of
Q55 1990–2012. The results confirm the existence of an N-shaped EKC relationship between income and
Q56 environmental degradation. The study offers a novel methodological contribution that makes it possible to
032
explain the environmental pollution process through the analysis of low-carbon technologies. This demonstrates
Keywords: how income levels affect energy consumption and how higher energy demand leads to a larger share of fossil
Environmental Kuznets curve sources in the energy mix and, thus, increased greenhouse gas (GHG) emissions. The effect on per capita GHG
Economic growth
emissions is explored in a model containing a dampening variable that moderates the relationship between
Energy RD & D
energy consumption and income. This empirical evidence helps to explain the interaction between energy
Renewable energy
regulation, economic growth and carbon emissions. This study also confirms the positive effect that energy
innovation process exerts on environmental pollution. Finally, it is noted that renewable energy sources help to
improve air quality.

1. Introduction
Although the Club of Rome raised awareness of the need for
economic systems to preserve environmental sustainability in the
1970s (Meadows et al., 1972), it was not until the early1990s that
environmental pollution problems began to be more frequently ad-
dressed in the theoretical economic literature (Grossman and Krueger,
1991; Shafik and Bandyopadhyay, 1992; Panayotou, 1993; Selden and
Song, 1994). These studies proposed the existence of an inverted-U
relationship between economic growth and environmental degrada-
tion, which was studied through the empirical hypothesis of the
environmental Kuznets curve (EKC). Increases in greenhouse gas
(GHG) emissions, a fundamental problem for economic development,
are encouraged by multiple variables, including economic growth,
economic structure, energy dependency and energy efficiency
(Velthuijsen and Worrell, 2002; Shi, 2003; Acaravci and Ozturk,
2010; Tugcu et al., 2012; Apergis and Ozturk, 2015; Ben Jebli et al.,
2016; Mensah, 2014; Al-Mulali et al., 2016).
In addition to the burden of environmental pollution, developed
economies also face the problem of energy dependency and high energy
intensity. Climate change can, in theory, be addressed by limiting
production that uses fossil fuels and improving the efficiency with
which such fuels are used. However, a key driver of emissions
reductions will be induced product and process innovation (Aghion
et al., 2014). Thus, not only are energy policy instruments associated
with the reduction of fossil energy sources, but public budget on energy
research, development and demonstration (RD & D), procedures to
improve energy efficiency, and measures intended to increase the share
of renewable sources in the energy mix have also become central in
environmental policy programmes. The present study used an EKC
model1 to examine the impact that economic growth, technology
innovation measures, and consumption of renewable energy sources
have on the evolution of per-capita carbon emissions in a selection of
17 OECD countries (Fig. 1).
The results of the regression, combined with economic growth,
make it possible to determine the effectiveness of these energy
measures in a context of sustainable economic growth.
Moreover, the results facilitate the analysis of the importance of

⁎
Corresponding author.
E-mail addresses: Agustin.Alvarez@uclm.es (A. Alvarez-Herranz), Daniel.Balsalobre@uclm.es (D. Balsalobre-Lorente), Shahbazmohd@live.com (M. Shahbaz),
Josemaria.cantos@uclm.es (J.M. Cantos).
1
Austria, Canada, Denmark, Finland, France, Germany, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States.

http://dx.doi.org/10.1016/j.enpol.2017.03.009
Received 29 November 2016; Received in revised form 2 February 2017; Accepted 3 March 2017
Available online 09 March 2017
0301-4215/ © 2017 Elsevier Ltd. All rights reserved.
A. Alvarez-Herranz et al.
40000
35000
30000
25000
20000
15000
10000
5000
0 13500000
1990 1994 1998
GDPpc
Fig. 1. Evolution of GHG emissions and GDPpc in OECD countries. Notes: Primary Axis: U$D GDP Per head, current prices, current PPPs. Secondary axis: GHG Thousands Tonnes of
CO2 equivalent.
Source: OECD (2016).
energy sector and the impact of shares of fossil and renewable energy
sources in the energy mix throughout the transition from developing
(low-income) to developed (high-income) economy (Wackernagel
et al., 1999; Turner and Hanley, 2011; UK Climate Change
Committee, 2008). Fossil sources are known to predominate in most
advanced countries’ energy mixes. In developing economies, energy
structure is based on over-exploitation of natural and fossil resources,
although in recent decades this model has been altered by the increase
in renewable sources and the implementation of innovations thought to
be conducive to a more sustainable model in energy sector.2 This new
scenario reflects a consensus on the need to increase environmental
sustainability through the use of low-carbon technologies. The im-
plementation and effectiveness of environmental regulations will play a
decisive role in thelong-term evolution of environmental pollution
levels (Bilgili et al., 2016; Balsalobre et al., 2016).
Nowadays energy regulation measures have taken on special
relevance in environmental correction processes; they are necessary
to correct the global warming process and help to ensure sustainable
economic growth (Dooley, 1998; Mendiluce et al., 2010; Costantini and
Martini, 2010; Dogan and Seker, 2016). In addition, because of
market-driven technological progress or government regulation eco-
nomic sectors may adopt less polluting technologies (Cole et al., 2005).
To this end, the importance of energy regulation policies for the
correction of environmental contamination processes must be clarified.
The main objective of this study was to demonstrate the existence of
EKC patterns for the selected countries and to show how energy
regulation measures, together with energy innovation and energy
substitution processes, impact GHG emissions.
2. Theoretical framework
The empirical environmental Kuznets curve (EKC) hypothesis
suggests the existence of a relationship between economic growth
and environmental pollution (Grossman and Krueger, 1991;
Panayotou, 1993; Selden and Song, 1994).
This section will discuss the basic aspects of theoretical behaviour
of the EKCmodel. This behaviour is reflected in the debate over the
fundamental determinants of long-term improvements in environment
correction, especially in the growing literature on the relationship
2
Energy practices have a direct impact on global warming processes. The global
average temperature has already increased over the last three decades and the global
surface temperature is likely to rise an additional 1.1–6.4°C (2.0–11.5°F) in the 21st
century (IPCC2007). Global increases in CO2 concentrations are primarily due to the
burning of fossil fuels, which accounts for 56.6% of total emissions. According to an IPCC
report, CO2 alone accounts for 76.7% of total GHG (IPCC 2007).
387
Energy Policy 105 (2017) 386–397
17000000
16500000
16000000
15500000
15000000
14500000
14000000
2002 2006 2010
GHG
between economic growth and pollution. The EKC postulate suggests a
relationship between environmental quality and economic growth, but
this relationship changes once a certain income threshold is attained.
In their pioneering work, Grossman and Krueger (1991) proposed an
inverted-U shaped relationship between environmental degradation
and income level. According to this theory, in the early stages of
economic growth, environmental pollution levels rise until reaching a
certain turning point, beyond which economies experience a reduction
in pollution levels (Fig. 2).
Fig. 2 shows an inverted-U shaped relationship between income
level and environmental pollution, which supposes a dynamic process
of structural changes connected with economic growth (Dinda, 2004).
This behaviour also implies that economic growth affects environmen-
tal quality through three channels (Grossman and Krueger (1991)):
a) Scale effect: when economic systems achieve a given level of
technology, increase in the inputs employed to obtain output, in
the early stages of economic growth, entails an increase in
environmental pollution levels. Specifically, the increased energy
requirements of the production function lead to greater use of fossil
sources and, consequently, increased pollution. The scale effect thus
decreases environmental quality (Torras and Boyce, 1998; Prieur,
2009).
b) Composition effect: the evolution of economic growth appears to be
linked to structural transformation, that is, the transition from a
developing economy, with highly polluting production processes, to
a more developed one, with a production pattern involving less-
polluting activities. In other words, the composition effect refers to
developing economies’ transition from capital-intensive industrial
sectors to service sectors and, ultimately, to technology-intensive
knowledge economies that employ more efficient energy procedures
and fewer fossil sources that directly impact environmental quality.
c) Technical effect: high-income (developed) economies allocate more
resources to RD & D. Fisher-Vanden et al. (2004) found evidence
that public budget on energy RD & D has a positive impact on
reducing energy intensity (consumption per unit of gross domestic
product) and, by extension, on the reduction of GHG emissions.
Developed economies replace old, dirty and inefficient technologies
with new, more efficient ones, thereby enhancing environmental
quality (Copeland and Taylor, 2004).
When the total effect of the relationship between economic growth
and environmental pollution is broken down, the technical effect is
considered to be the main factor in the correction of environmental
pollution process (Andreoni and Levinson, 1998; Markandya et al.,
2006). In this regard, technical obsolescence will lead to re-emergence
A. Alvarez-Herranz et al.
Fig. 2. EKC: Inverted-U shaped relationship between environmental pollution and
income level.
Source: Prepared by the authors.
of increasing pollution levels once the scale effect exceeds the composi-
tion and technical effects (Balsalobre and Álvarez, 2016). In keeping
with this point, Heerink et al. (2001) consider that the extent to which
the technical effect dominates the total effect depends on the incentives
for policy makers. In this regard, in developing countries, the scale
effect implies increased pollution levels, whilst in developed countries,
with lower growth rates but high income levels, where decontamination
processes depend mainly on innovations in low-carbon technologies,
technological changes could possibly offset the scale effect and delay
technical obsolescence (Andreoni and Levinson, 1998; Aghion et al.,
2014).
If one considers the environmental pollution process to be the
result of the scale, composition and technical effects, then as an
economy's income level increases, society will tend to demand greater
progress on clean environment policies aimed at protecting the
environment.3 The N-shaped pattern of the EKC makes it possible to
discuss aspects related to the scale effect and the long-term effects of
energy innovation. This, in turn, makes it possible to analyse the
potential return to rising emission levels once economies have achieved
negative pollution rates, and environmental technical obsolescence
becomes possible (Balsalobre and Álvarez, 2016). Fig. 2 does not reflect
such behaviour; thus, the pattern suggests a different behaviour, better
reflected by an N-shaped EKC model, where the return to rising
pollution levels occurs once the economy has achieved long-term high
income levels. The N-shaped behavioural pattern (Fig. 3) helps to
explain how economies can correct their technical obsolescence by
implementing long-term energy regulation policies.
The N-shaped EKC suggests a behaviour that amplifies the income-
environmental pollution relationship in the long term (Shafik and
Bandyopadhyay, 1992; Selden and Song, 1994; Grossman and Krueger,
1995; Moomaw and Unruh, 1997; Torras and Boyce, 1998). This
amplified relationship happens when the relationship between the
environmental degradation of the economies and income level is
initially positive, but becomes negative once a given income threshold
is reached, before ultimately becoming positive again. In other words,
this behaviour means that environmental degradation first increases in
an initial low-income stage (developing economies), then decreases
after a given income threshold is reached, and finally begins to increase
again, in a third stage marked by high income levels, but low economic
growth rates.
Therefore, the theoretical explanations of the EKC imply an
automatic decoupling between economic growth and environmental
3
This condition implies that economies will increase their innovation measures to
avoid energy obsolescence, in accordance with increased scale returns entailing an
elasticity of demand for a clean environment that exceeds unity (Dinda, 2004). On the
other hand, de Bruyn and Heintz (1999) suggest that whilst information on the
relationship between income and demand for environmental services is relevant to
explaining the shape of the curve, it is not enough to fully explain it on its own.
388
Energy Policy 105 (2017) 386–397
Fig. 3. EKC: N-shaped relationship between environmental pollution and income level.
Source: Prepared by the authors.
degradation. This process is based on the following assumptions: first,
pollution-related marginal disutility increases with income; second, the
availability of abatement technologies increases with economic devel-
opment; and third, the existence of an efficient institutional system
guarantees that the economy will follow neoclassical reasoning. On the
other hand, the EKC model is often criticised for the large sensitivities
frequently registered amongst EKC studies, which report very differ-
ently shaped EKCs depending on the selected time period or country
samples (Grossman and Krueger, 1991; Selden and Song, 1994) or the
existence of omitted variables. The present study attempts to address
the identified problems through the use of a data panel for a number of
countries in order to compensate for the existence of omitted variables
through corrections for endogeneity.
In order to demonstrate the hypothesis of this study, it must be
accepted that once an economy achieves a high level of income, society
will demand improvements in environmental quality processes that
will require efforts in the form of environmental protection measures
(Bruvoll et al., 2003). The resulting regulatory policies will increase the
technical effect through more efficient and less contaminant energy
production actions (Turner and Hanley, 2011).
Hypothesis 1:. Economic growth changes the shape of the EKC: in
an initial stage, increases in the income level entail higher GHG
emission levels; in a second stage, further increases in the income level
lead to a reduction in GHG levels; finally, in a third stage, the
economy begins to experience increases in technical obsolescence, at
which point a positive relationship once again emerges between
income level and GHG emission levels.
In addition to verifying an N-shaped EKC pattern for selected
OECD countries, this study attempts to demonstrate how, in the
absence of energy regulation policies, economies reach technical
obsolescence sooner, i.e. how energy regulation policies are the key
to delaying scale effect. In keeping with Torras and Boyce (1998), when
economies begin to push their technological limits, they experience a
return to a path of rising pollution levels due to a scale effect that
overshadows the joint impact of composition and technical effects. This
occurs once potential improvements in technological efficiency have
been exhausted or become prohibitively expensive (de Bruyn et al.,
1998). Gangadharan and Valenzuela (2001) showed that in the early
stages of development, the environment is largely neglected, but once a
certain income threshold is reached, society begins to demand a cleaner
environment, such that the environment can be considered a luxury
good. This relationship is not as automatic; thus, for this problem,
government reactions generally consist of implementing an induced
policy response, where it is necessary to take advantage of actions
linked to environmental regulation (Grossman and Krueger, 1995;
Balsalobre et al., 2015).
This study investigates the role of the promotion of low-carbon
technologies in correcting environmental pollution levels. To this end,
A. Alvarez-Herranz et al.
it is necessary to examine an additional variable, namely, the share of
renewable energy consumption, in order to determine the relationship
between economic growth and energy consumption. This relationship
has been the subject of basic research on energy in the economic
literature; however, no consensus has been reached in terms of results
(Grossman and Krueger, 1995; Halicioglu, 2009).
This notwithstanding, the relationship between economic growth
and energy consumption is not new in the economic literature. Several
studies consider energy consumption in the context of four growth
hypotheses (growth, conservation, feedback and neutrality)4 that
support the idea of interdependence between energy consumption
and economic growth (Shahbaz et al., 2012; Alper and Oguz, 2016):
Additional studies have considered the link between energy con-
sumption and economic growth, although there is no consensus in the
energy literature regarding a causal relationship between them. Other
studies have analysed the impact of energy sources on economic
growth (Bowden and Payne, 2010). Another approach was that taken
by Balsalobre and Álvarez (2016), who found, in relation to a selection
of OECD countries, that energy consumption was linked to economic
growth through the construction of a dampening variable, which has a
direct effect on GHG emissions. In other words, energy demand is
linked to economic growth, and this connection affects air pollution
levels. The present study incorporates this hypothesis, seeking to fill
the gap in the research by applying corrections for endogeneity and
inverted-V lag delays.
Having looked at several studies analysing the link between energy
consumption and economic growth, we will now highlight different
studies showing the existence of a relationship between energy
consumption and environmental degradation. Many researchers have
shown interest in factors associated with energy use that affect GHG
emissions. Ang (2007) demonstrated that energy consumption has a
direct impact on environmental pollution levels. Lean and Smyth
(2010) found a positive, long-term relationship between electricity
consumption and GHG emission levels. In another study, Halicioglu
(2009) noted that GHG emissions are determined by energy consump-
tion and economic growth. All of these studies show the positive impact
that energy consumption has on increases in GHG emission levels.
Recent literature shows evidence about the nexus between income, air
pollution levels and energy consumption (Sadorsky, 2009; Menyah and
Wolde-Rufael, 2010; Apergis and Payne, 2014, 2015; Sebri and Ben-
Salha, 2014; Dogan and Seker, 2016, among others). Apergis and
Payne (2015) analyse the link between renewable energy consumption,
income and CO2 emissions. They find a positive relationship renewable
energy consumption for sustainable economic growth. In addition,
Dogan and Seker (2016) explore the relationship between CO2 emis-
sions, renewable and non-renewable energy consumption, and the
income levels for European Union nations during 1980–2012. They
conclude that increases in renewable energy consumption reduce CO2
emissions, while the intensification in non-renewable energy consump-
tion enhances environmental pollution.
Drawing on these theories, our study proposed an additional
explanation based on the connection between energy consumption,
4
The growth hypothesis proposes the unidirectional causality running from energy
consumption to economic growth. Thus, the decrease in energy consumption has a
negative impact on economic growth (Ewing et al., 2007; Sari and Soytas, 2007; Payne,
2010a, 2010b, 2010c). Magnani and Vaona (2013). The conservation hypothesis
supports the existence of unidirectional causality running from economic growth to
energy consumption. In this case, reducing energy consumption will not affect economic
growth adversely (Sadorski, 2009 Ben Jebli and Ben Youssef, 2015b; Ozcan, 2013). The
feedback hypothesis considers a bidirectional causality between energy consumption and
economic growth. This relationship shows that reducing energy consumption has a
negative impact on economic growth and vice versa (Apergis and Payne, 2010a, 2010b,
2011, 2012b; Sadorsky, 2009, 2011, 2012; Tugcu et al., 2012). Finally, the neutrality
hypothesis provides for causality between economic consumption and economic growth,
whereby reducing energy consumption does not adversely affect economic growth
(Payne, 2009; Menegaki, 2011).
389
Energy Policy 105 (2017) 386–397
economic growth and environmental pollution variables. To validate
this hypothesis, we considered renewable energy consumption to
interact with economic growth and affect GHGpc emission levels.
This supposition suggests the existence of a dampening effect both
between economic growth and renewable energy, and on GHG emis-
sions.
We assumed that the study's dampening effect was a suppression
effect reflecting how the consumption of renewable sources in the
energymix is moderated (partially suppressed) by the economic cycle,
and how this negatively affects the overall impact of renewable sources
on the correction of GHG emission levels. In other words, this study
specifically sought to examine the effect of renewable energy usage on
pollution levels, whilst at the same time linking the interaction between
income levels (GDPpc) and renewable sources (RNW) to the correction
of GHGpc emissions.
It would also be appropriate to consider the need to increase the
share of renewable energy sources in the energy mix in order to reduce
the negative effect of overall energy demand on an ascending economic
cycle involving an increase in GHGpc emission levels. Several studies
have demonstrated that changes in the energy-mix pattern and the
promotion of renewable energy sources have a direct impact on the
reduction of GHG emission levels (Balsalobre et al., 2016).
When economies implement energy replacement processes to
increase renewable energy sources, the regulatory policies are limited
by the degree to which society is able to internalise environmental
pollution costs (Alfranca, 2009). Assuming that any environmental
regulation involves an initial cost that society must assume (Agostini
and Padilla, 2010), the economic cycle is linked to the promotion of
renewable energy sources and the isolation of environmental actions.
Under a scenario of high economic growth (developing economy stage),
the environmental correction process may be deficient without im-
provements in energy RD & D (Sevilla et al., 2013). Thus, in the short
term, the promotion of renewable energies can negatively impact the
final energy cost and, therefore, the economic system as a whole,
regardless of its positive contribution to the decarbonisation process
(Del Rio et al., 2009). In other words, the consequences of innovation
in the early stages of economic growth can be slow and difficult to
visualise by the economies, especially given the uncertainty surround-
ing their ultimate effects (Wang et al., 2012a, 2012b; He and Zhang,
2012). Another premise of this study is that technological progress
helps to improve environmental quality and, by extension, that the
technology effect is one of the main drivers able to delay the return to
an ascending stage in the N-shaped EKC pattern (Fig. 3). In this regard,
technological progress can help to replace fossil energy sources with
renewable ones and high-efficiency technologies in order to reduce air
pollution levels. When economies stimulate energy innovation pro-
cesses, it gives rise to reductions in both the level of energy intensity
and air pollution levels (Lin and Polenske, 1995). On the other hand,
the implementation of energy technologies may initially cause hazar-
dous new costs whilst increasing productivity (Fig. 4).
Several studies have found evidence that energy innovation posi-
tively impacts the control of GHG emission levels (Shafik and
Bandyopadhyay, 1992; Andreoni and Levinson, 1998). These studies
also provide evidence of this within the EKC model, concluding that
economic growth alone is insufficient to resolve environmental pollu-
tion problems.
Romer (1990) showed how innovation externalities can reduce
environmental pollution problems, which, in turn, can generate
innovation and the spread of acquired knowledge economies, making
technological innovation analysis essential in the long-term study of
environmental pollution process. Accordingly, investments in pollution
abatement equipment can improve environmental quality, especially
with regard to energy-related pollutants such as GHGpc. Consequently,
this theoretical approach suggests that energy innovation processes
help reduce environmental and social externalities. This supposition
makes it possible to validate that the returns on private investment in
A. Alvarez-Herranz et al.
Fig. 4. Turning points of the EKC model with regulatory variables and endogeneity
correction. Notes: The coefficients β1 > 0, β2 < 0, and β3 > 0 indicate a cubic polynomial in
an N-shaped EKC. The coefficients β1, β2, and β3 also allow us to calculate the turning
points in the cubic EKC model. X(1) represents the first turning point and X(2) the
second.
RD & D are lower than the social benefits, which, without public
intervention, would be difficult to achieve (Griliches, 1992). If one
assumes that environmental pollution is a negative externality, then the
empirical evidence suggests a relationship between low-carbon tech-
nological innovations and the reduction of environmental pollution
(Heyes and Kapur, 2011; Balsalobre et al., 2015.). Furthermore,
innovation achievements aimed at environmental correction measures
are premised on the idea that the expansion of clean technologies will
promote a reduction in environmental pollution levels (Aghion and
Howitt, 1992). This supposition implies increasing long-term returns,
which promote economic growth with decreasing levels of environ-
mental pollution. Therefore, sustainable economic growth involves
constant or increasing returns to scale, something that is supported
by energy innovation activities (Torras and Boyce, 1998; Andreoni and
Levinson, 2001).
Recent studies have confirmed this positive interaction between
advances in energy processes and the correction of environmental
quality within the EKC framework (Dinda et al., 2004; Cantos and
Balsalobre, 2013; Balsalobre and Álvarez, 2016).
Consequently, regulatory policies in the energy sector (introduction
of taxes or law enforcement) have proven effective in environmental
correction processes (Vehmas, 2005; Kumbaroglu, 2003; Costantini
and Mazzanti, 2012). In light of these findings, the present study aims
to verify whether public budget on energy RD & D has a positive effect
on environmental correction.
Hypothesis 2:. The energy regulation process reduces GHG
emission levels; energy innovation developments delay thelong-term
return to increasing pollution levels.
Hypothesis 3:. When economies incorporate both renewable energy
sources and energy innovation processes, these measures reduce the
income threshold that must be met for the long-term reduction of GHG
emission levels.
Hypothesis 4:. In the early stages of development, the
implementation of energy regulation policies involves a higher
income threshold, because the implementation of these measures
entails a cost that societies have to assume.
3. Empirical model
Behind the theoretical scheme outlined in Grossman and Krueger
(1995), which links a relationship between environmental pollution
390
Energy Policy 105 (2017) 386–397
and income levels to an N-shape,5the EKC's cubic scheme can be
expressed as follows:
EPit = α i + β1GDPpcit + β2GDPpc2it + β3GDPpc3it + β4Z it + εit (1)
EPit is environmental pollution of country i in year t, GDPpc is
income level per capita, and Zit determines additional variables that
impact environmental pollution. The coefficient αi accumulates envir-
onmental pressure when the average income level is of no particular
relevance in country i in year t. The β coefficients represent the relative
importance of exogenous variables, and εit is the error term, which is
normally distributed with zero mean and constant variance.
This study sought to fill the gap in the research regarding the
existence of a suppressor effect exerted by economic growth on the
impact that renewable sources have on the correction of GHG emission
levels. In addition, the variable inverted-V lag, proposed by De Leeuw
(1962), incorporates delays in the overall effect of RD & D on GHGpc
levels. In other words, this study incorporates the auxiliary variables
related to public budget on energy RD & D (through a De Leeuw, or
inverted-V, scheme of finite delays) and the percentage of renewable
energy consumption in the energy mix. These additional variables
facilitate analysis of the role of energy regulation processes in the
evolution of GHG emission levels.6 To validate this hypothesis, we built
Eq. (2):
GHGpcit = α i + β1GDPpcit + β2GDPpc2it + β3GDPpc3it + φ1RNWit
+ φ2GDPpcit *RNWit + δ0RDETpcit + δ1RDETpcit−1
+ δ 2RDETpcit−2 + δ3RDETpcit−3 + δ4RDETpcit−4 + εit (2)
⎧ (j + 1)δ 0 ≤ j ≤ s /2
where δj = ⎨
⎩ (s−j + 1)δ s /2 + 1 ≤ j ≤ s
The estimation of a lag distribution model of Eq. (2) faces two
challenges. First, each new delay included reduces the model's freedom
and, thus, the accuracy of the estimates. Second, since the reference
variable appears as an explanatory variable at different times, the
model can exhibit multi-colinearity.
It is thus necessary to transform Eq. (2) into Eq. (3) in order to
eliminate the problems of a lag distribution model:
GHGpcit = α i + β1GDPpcit + β2GDPpc2it + β3GDPpc3it + φ1RNWit
+ φ2GDPpcit *RNWit +δ ZDit + εit (3)
where:
s
s=4
2
ZDit = ∑ (j + 1)RDETpcit−j + ∑ (s − j + 1)RDETpcit−j
j=0 j= s +1
2
where:
1. GHGpcit is per capita GHG emissions (million tonnes of CO2
equivalent) for country i in year t (OECD, 2016).
2. GDPpcit is income level in per capita terms, in millions of dollars in
purchasing power parity ($US million, current PPPs) for country i
5
The behavioural patterns the EKC may follow depend on the value of the β
coefficients in Eq. (1). Of these various patterns, we will pay special attention to the
N-shaped one.
a)
If β1 > 0, β2 < 0, and β3=0, a quadratic relationship in an inverted-U shaped pattern
indicates that high income levels are associated with decreasing levels of pollution
beyond a certain income threshold.
If β1 > 0, β2 < 0, and β3 > 0, a cubic polynomial reveals an N shape, such that the inverted-
U hypothesis holds up to a certain point, beyond which pollution increases again.
6
Different studies have found evidence of a positive relationship between low-carbon
technologies and the decontamination process in which technological innovation is
essential for the process (Anderson and Cavendish 2001; Pizer and Popp 2008; Cantos
and Balsalobre, 2013).
A. Alvarez-Herranz et al.
and year t. In its cubic expression, it verifies that the EKC is N-
shaped for the analysed countries over the period (OECD, 2016).
3. RDETpcit is the public budget for energy research development and
demonstration (RD & D) in per capita terms (US$ millions current
PPPs) for country i in year t (OECD, 2016).
4. RNWit is the share of renewable energy consumption in the energy
mix for country i in year t (OECD, 2016). Braun et al. (2010) found
that renewable technologies create knowledge spillovers at the
national level.
5. GDPpcit*RNWit is the dampening variable reflecting the magnitude
and/or direction of the relationship between the independent
variable (RNWit) and the response variable (GHGpcit), by amplifying
or even inverting (suppression effect) the causal effect. We used
moderation effects to prove the causal hypotheses (Wu and Zumbo,
2008).
6. ZDpcit-j is the public budget for energy RD & D in per capita terms
($US, current prices, current PPPs) in country i over period t-j,
where j=1, 2, 3, 4 corresponds to time lag. This variable reflects the
delay in t-j periods, which is incorporated in Eq. (3). It has a
structure of finite delays of the fourth order, forming a finite
inverted-V-shaped lag (De Leeuw, 1962). This variable contains
the multiplier effect of the variable RDETpcit-j on the endogenous
variable GHGpcit, which increases until reaching its maximum
intensity at the j=2 value, after which its intensity begins to decline
(OECD, 2016).
Despite the extensive literature investigating the EKC hypothesis,
there is a lack of research incorporating innovation delay7
(Dechezleprêtre, 2011; Aghion, 2014). In the present study, the energy
innovation variable ZDpcit-j examines the effect of delays on the
environmental pollution process. This contribution helps to enrich
the literature on the EKC.
Although, the variables RNWit, GDPpcit*RNWit and ZDit are
considered exogenous, in the present study we considered GDPpcit to
be an endogenous explanatory variable that appears to be correlated
with the error term εit. We used the Hausman test to check for the
endogeneity of the GDPpcit, GDPpc2it and GDPpc3it variables.
In the first stage of the econometric estimation procedure, Eq. (3) is
estimated by panel least squaring (PLS) to find the reduced form of the
endogenous explanatory variable based on the exogenous variables and
possible instrumental variables.
In order to mitigate the problems of endogeneity, it was necessary
to incorporate an instrumental variables approach in the regressions
both with and without fixed effects to identify the coefficient of GDPpc.
The incorporated instruments were as follows:
• AGEDRit is the age dependency ratio (% of working-age population)
in country i and year t (World Bank, 2015). The higher the age
dependency ratio is, the lower the rates of growth and GDPpc, both
because countries with large populations of young people are likely
to be less productive on average and because poorer countries tend
to have this demographic profile (World Bank, 2015).
• URBPit is the percent of urban population in the total population of
country i. URBPit represents the share of people living in urban
areas. The data were collected and smoothed by the United Nations
Population Division. Bruno and Easterly (1998) and Anwar and Sun
(2011) empirically tested the impact of urban population on
economic growth and showed how this variable has a statistically
significant influence on economic growth.
Therefore, AGEDRit and URBPit are plausible and appropriate
7
Dechezleprêtre et al. (2013) found that spill-overs from low-carbon innovation are
over 40 per cent greater than with conventional technologies (in the energy production
and transportation sectors).
391
Energy Policy 105 (2017) 386–397
instruments for GDPpcit (Lin and Liscow, 2013). These instruments
are correlated with GDPpcit, whereas they did not affect the quality of
GHGpcit, except through their effect on GDPpcit.
4. Results and discussion
Having explained the theoretical model, we will now estimate and
analyse the econometric results obtained from Eq. (3) in order to verify
the effect that, together with economic growth, the variables RNWit,
GDPpcit*RNWit and ZDit have on GHGpc emissions levels within the
EKC framework.
Eq. (3) is estimated as a fixed-effects panel data model, which is
appropriate if there is unobserved heterogeneity in specific countries.
To estimate the econometric model proposed in Eq. (3), we used the
panel least squares (PLS) method. This method is suitable when the
source of the dependent variable as individual heterogeneity, unobser-
vable and biases caused by faulty specification (Table 1).
Eq. (3) was estimated by PLS specification, assuming cross-section
heteroscedasticity and autocorrelation through the incorporation of an
AR(1) structure. It was necessary to confirm whether the data
supported the main hypothesis, based on the notion that GHGpc
emission levels are influenced by economic growth (GDPpc). The
regression (Table 2) also revealed the existence of a dampening effect
between income levels and renewable energy sources
(GDPpcit*RNWit), which constitutes a methodological advance for the
EKC model.8 This explains the relationship between the economic
effort countries make in relation to budget on energy RD & D and the
dampening effect it has on energy intensity. When a country makes a
large effort in energy innovation, it causes a reduction in its rates of
energy intensity. Indeed, once a dampening effect on the causal
relationship between an independent variable X (RNWit) and endo-
genous variable Y (GHGpcit) is detected, it is necessary to show the role
of a third variable, Z, called a ‘suppressor’ (GDPpcit).
The estimation results of Eq. (3) revealed the existence of specific
individual effects in each country affecting its decisions. If the model
does not consider these latent effects, there will be a problem of
omitted variables and the explanatory variable estimators will be
biased. Therefore, the next step of the study was to check for the
existence of endogeneity (Table 3). The existence of any endogenous
explanatory variable in Model 1 implies that the PLS method was
inconsistent, making it necessary to apply the instrumental variables
method (Two-Stage Least Squares (TSLS)), which is unbiased and
consistent. In order to mitigate the endogeneity, it was necessary to
restructure the model (Eq. (3)), using instrumental variables without
fixed effects to determine the income coefficient.
The instrumental variables must be sensibly reliable and correlated
instruments for GDPpcit, but they only affect GHGpcit through their
effect on GDPpcit. For this study, the exogenous variables URBPit and
AGEDPit were considered instruments for the variables GDPpci,
GDPpc2it and GDPpc3it, making it necessary to verify whether these
instruments are individually and jointly significant in Eqs. (4)–(6) up
to a reasonably small significance level (not more than 5%), as can be
seen in the t-statistic and Wald test (Tables 3 and 4).
GDPpcit = π 0 + π1RNWit + π 2ZDit + π3RNWit *GDPpcit
+ π4URBPit + π5AGEDR it + π6URBP 2it + π 7AGEDR 2it
+ π8URBP 3it + π 9AGEDR 3it + V1it (4)
8
The hypothesis of moderation, the effects of which are also called interaction effects,
seeks to determine under what conditions a relationship becomes stronger or weaker,
disappears or changes direction. The moderator variable may be qualitative or
quantitative and affects the magnitude and/or direction of a relationship between an
independent variable (predictor) and dependent variable (criterion) (Cohen 2003).
A. Alvarez-Herranz et al. Energy Policy 105 (2017) 386–397

Table 1
Variable statistics.
Sources: World Bank (2015)

Stat. GHGpc GDPpc RDETpc RNW ZD URBP AGEDR

Mean 0.012342 31,711.06 13.41.406 34.34104 111.6747 76.95545 50.01401

Median 0.011143 30,518.63 11.25.773 25.38125 98.87453 78.34000 49.71716
Maximum 0.025288 66357.73 96.71.398 99.71511 687.7149 91.90200 60.02857
Minimum 0.006051 12,901.31 0.174567 1.579086 2.852430 50.44900 43.42873
Std. Dev. 0.004843 9188.215 12.22737 27.82472 89.00779 7.921405 3.057995
Skewness 1179.892 0.752810 2.671.512 0.753318 1.868069 −1.033410 0.271873
Kurtosis 3.549.441 3.810430 14.97820 2.483796 10.24433 4.071069 2.589907
Jarque-Bera 7.900668 3.934795 2315.172 34.13599 894.1.581 72.92978 6.242450
Probability 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.044103
Sum 3.986568 10,242,672 4332.740 11,092.16 36,070.91 24,856.61 16,154.53
S.Sq. Dev. 0.007552 2.72E+10 48,141.78 249,297.2 2,551,009 20,205.07 3011.130
Observations 323 323 323 323 323 323 323

Table 2
Estimation of Model 1 (Eq. (3)) without correction of endogeneity by Panel Least Squares
(PLS).
Table 3
Estimation of GDPpc regressions in Eqs. (4), (5) and (6) by Panel Least Squares (PLS).
Dependent variable: GHGpc
Sample (adjusted): 1994–2012
Dependent variables: GDPpc, GDPpc^2 and GDPpc^3
Cross-sections included: 17
Sample (adjusted): 1994–2012
Cross-sections included: 17
Model 1
White cross-section standard errors & covariance (no d.f. correction)

Variable Coefficient
Variable Dependent Var . Dependent Var. Dependent Var.
GDPpc GDPpc^2 GDPpc^3
C 0.013173*
[7.525326]
C 1163855* 4.13E+10** −8.77E+14***
(0.0000)
[3.643119] [2.275402] [−0.895008]
GDPpc 3.06E−07**
(0.0003) (0.0236) (0.3715)
[2.192042]
RNW −56.05025** −7535456* −7.34E+11*
(0.0292)
[−1.713887] [−2.907444] [−3.614272]
GDPpc^2 −1.05E−11**
(0.0876) (0.0039) (0.0004)
[−2.925549]
GDPPC*RNW 0.010536* 830.8296* 52290080*
(0.0037)
[36.87309] [29.27356] [18.77915]
GDPpc^3 9.36E−17*
(0.0000) (0.0000) (0.0000)
[3.114102]
ZD 1.135701 568975.5* 7.03E+10*
(0.0020)
[0.289850] [2.719250] [5.045115]
RNW −7.79E−05*
(0.7721) (0.0069) (0.0000)
[−7.767011]
AGEDR −40564.89** −1.26E+08 2.17E+14*
(0.0000)
[−1.930596] [0.122495] [4.052573]
GDPpc*RNW 3.78E−10
(0.0545) (0.9026) (0.0001)
[1.272645]
URBP −23400.67* −2.01E+09* −1.25E+14*
(0.2042)
[−4.303870] [−5.115069] [−5.287314]
ZD −5.92E−06*
(0.0000) (0.0000) (0.0000)
[−5.681662]
AGEDR^2 793.9507** −3397657 −4.34E+12*
(0.0000)
[1.877606] [−0.166411] [−4.120502]
AR(1) 0.571161*
(0.0614) (0.8679) (0.0000)
[10.92593]
URBP^2 355.8364* 29394240* 1.80E+12*
(0.0000)
[4.515151] [5.150993] [5.225289]
Effects specification: cross-section fixed (dummy variables)
(0.0000) (0.0000) (0.0000)
R-squared 0.992913
AGEDR^3 −5.185795** 26090.76 2.87E+10*
Adjusted R-squared 0.992335
[−1.832042] [0.194754] [4.178198]
F-statistic 1717.869
(0.0679) (0.8457) (0.0000)
Durbin-Watson stat 2.13718
URBP^3 −1.660014* −134817.4* −8.24E+09*
[−4.446571] [−5.010385] [−5.070210]
Notes: t-statistic and p-value are given in [ ] and ( ) respectively; *, **, *** show (0.0000) (0.0000) (0.0000)
significance at 1%, 5% and 10% respectively. Effects specification: cross-section fixed (dummy variables). Weighted statistics
R-squared 0.900898 0.921950 0.918036
Adjusted R- 0.892556 0.915380 0.911137
squared
F-statistic 107.9962 140.3302 133.0623
GDPpc2it = π 0 + π1RNWit + π 2ZDit + π 3RNWit *GDPpcit Prob(F-statistic) 0.000000 0.000000 0.000000
Durbin-Watson 0.177305 0.209232 0.233657
+ π4URBPit + π5AGEDR it + π6URBP 2it + π 7AGEDR 2it stat
+ π8URBP 3it + π 9AGEDR 3it + V2it (5)
Notes: t-statistic and p-value are given in [ ] and ( ) respectively; *, **, *** show
significance at 1%, 5% and 10% respectively.

392
A. Alvarez-Herranz et al. Energy Policy 105 (2017) 386–397

Table 5
Table 4 Estimation of Eq. (7) by Panel Least Squares (PLS).
Wald Test: Eqs. (4), (5) and (6).
Dependent variable: GHGpc
Eq. (4) GDPpc Eq. (5) GDPpc^2 Eq. (6) GDPpc^3 Sample (adjusted): 1994–2012
Cross-sections included: 17
Test statistic Value Value Value
df df df Variable Coefficient
Prob Prob Prob
F-statistic 216.2400* 198.5399* 135.5907* C 0.015064*
(6, 297) (6, 297) (6, 297) [10.81055]
(0.0000) (0.0000) (0.0000) (0.0000)
Chi-square 1297.440* 1191.239* 813.5439* GDPpc 1.67E−07***
6 6 6 [1.383980]
(0.0000) (0.0000) (0.0000) (0.1674)
GDPpc^2 −5.84E−12***
[−1.513065]
Notes: Null hypothesis: C(5)=C(6)=C(7)=C(8)=C(9)=C(10)=0; Wald test validates the
(0.1313)
instrumental variables. Notes: t-statistic and p-value are given in [ ] and ( ) respectively;
GDPpc^3 3.82E−17***
*, **, *** show significance at 1%, 5% and 10% respectively.
[0.964113]
(0.3358)
RNW −0.000113*
[−11.91909]
(0.0000)
GDPpc3it = π 0 + π1RNWit + π 2ZDit + π 3RNWit *GDPit
GDPpc*RNW 1.08E−09*
+ π4URBPit + π5AGEDR it + π6URBP 2it + π 7AGEDR 2it [6.074059]
(0.0000)
+ π8URBP 3it + π 9AGEDR 3it + V3it (6) ZD −4.81E−06*
[−5.532292]
It is now necessary to check that the URBPit and AGEDPit variables (0.0000)
are instruments of the GDPpci, GDPpc2it and GDPpc3it variables V1B 3.15E−07**
(Table 3): [2.287696]
(0.0229)
Table 3 shows the estimation of the residues of Eqs. (4)–(6), which V2B −8.99E−12**
have been renamed: RESZD1b by (V1t), RESZD2b by (V2t) and RESZD3b [−2.226539]
by (V3t). Once the three variables V1t, V2t and V3t were obtained, they (0.0267)
were entered in the PLS estimated equation to check for the existence V3B 8.66E−17**
[2.161868]
of endogeneity with regard to GDPpcit, GDPpc2it and GDPpc3it.
(0.0314)
To capture the unobservable effects specific to each country that do Effects specification: cross-section fixed (dummy variables)
not vary over time, a fixed-effects regression method was used, R-squared 0.990187
implementing GDPpc, GDPpc2 and GDPpc3 with regard to AGEDRit Adjusted R-squared 0.989361
dependence and the level of URBPit, including both the square and F-statistic 1198.709
Prob(F-statistic) 0.000000
cubic expressions of these instruments. The estimation results pro-
Durbin-Watson stat 0.965286
vided in Table 3 show that there was no correlation between the
instrumental variables for Eqs. (4)–(6) and the error term in Eq. (3). Notes: t-statistic and p-value are given in [ ] and ( ) respectively; *, **, *** show
The explanatory variables GDPpcit, GDPpc2it and GDPpc3it will not be significance at 1%, 5% and 10% respectively.
correlated with the error term (εit), if and only if the error terms V1it,
V2it, and V3it are uncorrelated with εit and incorporate the Figure dues
(X(1)=US$21,454.06). Beyond this point, higher income levels are
of Eqs. (4)–(6). To verify this lack of correlation, we included these
inversely related to GHGpc levels (GHGpc levels start to decrease) until
error terms in a second step and estimated Eq. (3), which became Eq.
GDPpc reaches a second turning point (X(2)=US$58,245.18) after
(7):
which GHGpc starts to increase again (Table 7).
GHGpcit = α i + β1GDPpcit + β2GDPpc2it + β3GDPpc3it + β4RNWit We will now analyse this EKC's behaviour and explain how
additional variables related to energy innovation processes and renew-
+β5GDPpcit *RNWit + β ZDit + δ1V̂1it + δ 2V̂2it + δ3V̂3it + εit able sources can positively impact the delay before the upturn in
6

(7)
We then estimated Eq. (7) by PLS and checked whether the
coefficients δ1, δ2, and δ3 ofV̂1it , V̂2it and V̂3it were statistically significant
(Table 5).
The results of Eq. (7) show that GDPpci, GDPpc2it, and GDPpc3it
are endogenous explanatory variables of the variablesV̂1it , V̂2it andV̂3it ,
which are statistically significant (Table 5). Thus, to solve the endo-
geneity problem of Eq. (3), we used the instrumental variables method
(TSLS) to obtain unbiased and efficient estimators (Table 6).
The coefficients β̂1 > 0, β̂2 < 0 and β̂3 > 0 determine the cubic shape of
the EKC. The behaviour of the remaining coefficients also helps explain
the relationship between income level and GHGpc emissions. The
result of the regression implies that, in an initial stage, increases in
income levels lead to increases in GHGpc emission levels until the first
turning point is reached.9
9
The formula used for the estimation of turning points for the cubic model was as
GHGpc levels. The negative coefficient β̂4 =−9.04E-05 in Eq. (3)
indicates an inverse relationship between the share of renewable
consumption in the energy mix (RNWit) and GHGpc emissions.
This result shows that the promotion of renewable sources is a
suitable energy policy tool that positively affects pollution correction
levels. The results of the regression also led us to consider that the
impact of renewable energy consumption on economic growth might
(footnote continued)
follows:
−β ± β 2 −3β β
Xj = 2 2 1 3
, ∀ j = 1,2
3β3
To estimate the turning points, it is necessary to make a change in coefficient β1, since the
breaking point where the function reaches the maximum and minimum value depends on
RNW. When the RNW variable appears in the moderate model GDPpc, it will affect the
first-order coefficient. Consequently, in the coefficient β1*= (β1 +δ3* RNW), the median
value of RNW is 2.538, justified by the asymmetric distribution of that variable.
Therefore, the model's breaking points have to be estimated based on a β1*=(4.96E-
07+1.02E-09*2.538)=4.99E-07 coefficient.

393
A. Alvarez-Herranz et al. Energy Policy 105 (2017) 386–397

Table 6 Table 7
EKC Model 1*(Eq. (3)) with correction of endogeneity by panel Two-Stage Least Comparison of EKC Model 1*(Eq. (3)) and EKC Model 2 (Eq. (8)) by TSLS.
Squares(TSLS).
Dependent variable: GHGPC
Dependent variable: GHGpc Sample (adjusted): 1994–2012
Sample (adjusted): 1994–2012 Cross-sections included: 17
Cross-sections included: 17 Instrument specification: C AGEDR URBP AGEDR^2 URBP^2, AGEDR^3,
Instrument specification: C AGEDR, URBP, AGEDR^2, URBP^2, AGEDR^3, URBP^3, RNW RNW*GDPPC and ZD.
URBP^3, RNW RNW*GDPPC and ZD
Lagged dependent variable & regressors added to instrument list Variable Model 1* Model 2

Model 1* C 0.011215* 0.001765

[5.612613] [0.238578]
Variable Coefficient (0.0000) (0.8116)
GDPPC 4.96E−07* 1.42E−06*
C 0.011215* [3.011435] [1.914230]
[5.612613] (0.0028) (0.0566)
(0.0000) GDPPC^2 −1.59E−11* −5.18E−11**
GDPpc 4.96E−07* [−3.628847] [−2.116076]
[3.011435] (0.0003) (0.0352)
(0.0028) GDPPC^3 1.33E−16* 5.42E−16**
GDPpc^2 −1.59E−11* [3.39857] [2.111409]
[−3.628847] (0.0008) (0.0356)
(0.0003) RNW −9.04E−05*
GDPpc^3 1.33E−16* [−8.07966]
[3.39857] (0.0000)
(0.0008) RNW*GDPPC 1.02E−09*
RNW −9.04E−05* [2.899367]
[−8.07966] (0.0040)
(0.0000) ZD −5.82E−06*
RNW*GDPpc 1.02E−09* [−5.267030]
[2.899367] (0.0000)
(0.0040) AR[1] 0.549118* 0.63652*
ZD −5.82E−06* [10.46096] [3.652343]
[−5.267030] (0.0000) (0.0003)
(0.0000) Effects specification: cross-section fixed (dummy variables)
AR[1] 0.549118 R-squared 0.992643 0.984165
[10.46096]*** Adjusted R-squared 0.992043 0.983054
(0.0000) F-statistic 1902.942 688.2462
Effects specification: cross-section fixed (dummy variables) Durbin-Watson stat 2.114916 1.554567
R-squared 0.992643 Second-Stage SSR 4.57E−05 0.000395
Adjusted R-squared 0.992043 Inverted AR Roots 0.55 0.64
F-statistic 1902.942
Durbin-Watson stat 2.114916 Notes: t-statistic and p-value are given in [ ] and ( ) respectively; *, **, *** show
Second-Stage SSR 4.57E−05 significance at 1%, 5% and 10% respectively.
Inverted AR Roots 0.55

Notes: t-statistic and p-value are given in [ ] and ( ) respectively; *, **, *** show
significance at 1%, 5% and 10% respectively.

depend on the structure and developmental stage of the economy (Saidi

and Hammami, 2015). Therefore, the Eq. (3) regression also revealed a
suppressor effect (β̂5=1.02E-09), reflecting the effect that income level
(GDPpc) has on renewable sources in the correction of GHGpc
emission levels. Although we believe that replacement with renewable
energy sources positively contributes to reducing GHG emission levels,
when a dampening effect is incorporated, it can be understood that in
an expansionary economic cycle, economies still rely on conventional
energy sources, and the net positive impact of renewable sources on
environmental correction declines. In other words, in stages charac- Fig. 5. Evolution of impact of RD & D: inverted-V lag (De Leeuw, 1962).
Source: Prepared by the authors.
terised by high energy requirements, economies with a high share of
fossil sources find it harder to control air pollution levels.
With regard to energy innovation, represented by the variable ZDit, the
GHGpcit = 0.011215 + 4.96E − 07 GDPpcit − 1.59E − 11 GDPpcit2
negative sign (β̂6 =−5.82E-06) suggests that an increase in public on energy
RD & D reduces GHGpc emission levels. The regression confirmed the + 1.33E − 16 GDPpcit3 − 9.04E − 05 RENEWABLit +
positive impact of public budget on energy RD & D on the GHGpcit 1.02E − 09 GDPpcit *RENEWABLit − 5.82E − 06 RDETpci t −0 − 1.16E
emissions of countries considered through different multipliers. Torras
− 05 RDETpci t −1 − 1.75E − 05 RDETpci t −2
and Boyce (1998) argued that the key to solving the scale effect is to
−1.16E − 05 RDETpci t −3 − 5, 82E − 06 RDETpci t −4 + εit
promote energy innovation (ZDit). When technical obsolescence occurs,
increases in income level will lead to increased environmental degradation Fig. 5 shows that the De Leeuw multiplier achieves its maximum
(scale effect) and a return to an upward trend in GHGpc emission levels. impact two years out. This multiplier helps to explain the impact that
It should be noted that the short-term multiplier, or impact, causes public budget on energy RD & D has on emission levels. The empirical
a change in the average value of GHGpcit emissions, which follows a evidence reveals that energy R & D measures have the greatest effect on
change in unitary public budget on energy RD & D. correcting pollution at lag 2. This confirms that energy innovation

394
A. Alvarez-Herranz et al.
Fig. 6. Comparison of models: Isolation of the energy regulation variables.
Source: Prepared by the authors.
Fig. 7. The effect of technological obsolescence in the EKC model.
Source: Prepared by the authors.
measures take two years to reach their fullest potential.
Lastly, it was necessary to isolate the effect of energy regulation in
the relationship between income level and air pollution (Model2),
omitting the variables ZDit and RNWit and, by extension,
GDPpcit*RNWit, to see how the EKC behaved (Eq. (8)). We then
compared the turning points adjusted for the omission of the regula-
tory variable. This step was significant to demonstrate the relevance of
energy regulation policies to solving environmental pollution problems
(Fig. 6)
GHGpcit = α i + β1GDPpcit + β2GDPpc2it + β3GDPpc3it + εit
A comparison of Model 1*(Eq. (3)) with Model 2 (Eq. (8)) shows
that, in an initial stage (Stage 1), the income requirements necessary to
achieve reductions in GHG emission levels are higher in Model 1 than
in Model 2 (X(1)=US$21,454.06 > X(3)=US$19,958.58). This implies
that when economies first apply energy regulation processes, society
has to assume an initial higher cost. Another consequence of imple-
menting energy regulation measures is that the income threshold for
the second turning point (Stage 2) and the return to increasing
pollution levels is higher when economies implement regulatory
improvements (X(2)=US$58,245.18 > X(4)=US$43,756.06). The sec-
ond stage indicates the effectiveness of energy-related regulatory
policies. Regulatory measures in the energy sector are partly justified
by delays in the long-term ascending pollution phase. In other words,
when economies implement regulatory policies in the energy sector, it
helps to prevent the scale effect and, thus, technical obsolescence
(Fig. 7).
One consequence of the results obtained in the regression of Eq. (3)
is that without energy innovation measures, technical obsolescence
forces the return to a stage of increasing environmental degradation.
Energy Policy 105 (2017) 386–397
Fig. 8. Conceptual scheme.
Source: Prepared by the authors.
Smulders and Bretschger (2000) built a growth model, within the EKC
scheme, that proposes that the relationship between economic growth
and environmental quality is the consequence of technological change,
sectoral shifts and changes in environmental regulation. In keeping
with these authors, we consider that the environmental correction
process requires substantial efforts in terms of energy innovation to
reorient the economic system towards less polluting sectors.
Consequently, to avoid a return to a path of increasing contamination,
energy regulation measures must be implemented that expand im-
provements in the energy sector with the aim of avoiding the trap of
decreasing technological returns on a path to technical obsolescence
(Fig. 8).
When governments implement approaches that seek to reduce
pollution levels, they must thus assume increasing returns to scale.
In turn, it can be concluded that technological innovation practices
make environmental correction possible at lower income levels.
Moreover, the implementation of measures to promote energy innova-
tions and renewable sources will result in a deviation from the
diminishing technological returns, thereby helping to reverse the
upward trajectory of the EKC (Torras and Boyce, 1998).
The positive value of the coefficient β5 (GDPpcit*RNWit) helps to
validate our hypothesis that energy consumption is linked to economic
growth, which implies a relationship with GHG emission levels.
Specifically, if an economy's energy system is dominated by fossil
(8) sources, then when the energy requirements increase, in keeping with
an expansive economic cycle, GHG emission levels will also increase,
because the share of non-renewable energy sources positively affects
GHG emission levels (Al-Mulali and Ozturk, 2016). In other words,
when economies undergo increased economic growth, energy demand
will increase, decreasing the share of renewable sources in the overall
energymix. Consequently, the key to solving this problem lies in
promoting renewable sources able to reduce the share of fossil sources
in the energymix.
Therefore, this study has shown that it is essential to increase the
presence of renewable energy sources in the energy-mix system in
order to reduce the impact of fossil energy sources on GHG emission
levels when the energy demand of an economy increases.
In keeping with the findings of this study, administrations should
thus implement regulatory policies, both to promote renewable sources
and with regard to energy innovation measures, to correct air pollution
levels. Such measures help to delay technical obsolescence and also
control the scale effect that drives economies to a return to increasing
pollution levels. Although the promotion of renewable sources has a
direct impact on the reduction of GHG emission levels in the short
395
A. Alvarez-Herranz et al.
term, in the longterm it is necessary to implement energy innovation
measures to delay technical obsolescence and, thus, the return to a
stage of increasing GHG emission levels.
5. Conclusions and policy implications
This study validates the relevance of energy regulation measures,
including both the promotion of renewable sources and energy
innovation, for the correction of GHG emission levels. The results of
the estimates distinguish amongst three elements that influence air
pollution correction: economic growth, the promotion of renewable
energy sources, and public budget on energy RD & D. The regression
results confirm the existence of a positive relationship between
improvements in energy innovation processes and the reduction of
GHGpc emission levels. In addition, when economies are at low-
income stage (developing countries), both the promotion of renewable
sources and public budget on energy innovation need to be increased
for achieving a reduction in GHG levels, even though it entails an added
cost that societies have to assume. Once economies reach a developed
stage, they have to continue increasing their energy regulation proce-
dures in order to delay the scale effect.
One policy implication of this study lies in the direct relationship it
reveals between increased public budget on energy RD & D and the
reduction of GHG emission levels. However, this effect requires a time
lag to become fully efficient. This finding confirms that energy
innovation efforts do not have a short-term effect; it is therefore
necessary to assume an initial cost that will yield its highest return in
the longterm.
Moreover, the process of replacing conventional energy sources
with renewable ones positively contributes to reducing GHGpc emis-
sion levels. This is also linked to the economic cycle, such that as an
economy's income level increases, the share of renewable energy in the
energy mix declines due to net increases in energy demand. This entails
an increase in the consumption of fossil sources to accelerate economic
growth that negatively affects air pollution levels. For this reason,
governments have to strengthen their efforts to promote renewable
sources to reduce energy dependency and the pollution impact of the
energy system when an economy's energy demands increase.
Finally, in analysing different variables, this study found the linking
of energy policies to innovation and the promotion of renewable
sources to play a positive role; in this regard, it may be useful for
policy makers to examine the effects of renewable sources and public
budget on energy RD & D on environmental pollution. This study
supports methodological improvements, as it is the first time that an
EKC methodology with correction of endogeneity has supported a De
Leeuw (1962), or inverted-V lag, within the model. This study also
considered the dampening effect of income level on the share of
renewable sources, showing that the economic cycle is closely linked
to the type of energy source used and that, when economies have a low
share of renewable sources, increased energy demand will have a
negative effect on GHG emission levels. Future empirical research
could seek to provide guidelines for energy and environmental policy
by incorporating other variables, such as corruption, water pollution or
trade openness, or to develop a model for individual countries, through
quantile analyses that help to explore the differences between devel-
oping and developed countries.
References
Aghion, P., Howitt, P., 1992. A model of growth trough creative destruction.
Econometrica 60 (2), 323–351.
Aghion, P., Hepburn, C., Teytelboym, A., Zenghelis, D., 2014. Path-Dependency,
Innovation and the Economics of Climate Change. Supporting Paper for New Climate
Economy. Grantham Research Institute on Climate Change and the Environment,
London School of Economics and political Science, London.
Agostini, L., Padilla, J., 2010. La promoción de las energías renovables: La experiencia
española, Electricidad Verde: energías Renovables y Sistema Eléctrico. In: Moselle,
396
Energy Policy 105 (2017) 386–397
B., Padilla, J., Schmalensee, R. (Eds.), Marcial Pons. Buenos Aires, Madrid,
Barcelona, 517–541.
Alfranca, O., 2009. Regulación ambiental e innovación. Clm. . Econ.: Rev. Econ. Castilla -
La Mancha 15, 9–31.
Al-Mulali, U., Ozturk, I., 2016. The investigation of environmental Kuznets curve
hypothesis in the advanced economies: the role of energy prices. Renew. Sustain.
Energy Rev. 54, 1622–1631.
Al-Mulali, U., Ozturk, I., Solarin, S.A., 2016. Investigating the environmental Kuznets
curve hypothesis in seven regions: the role of renewable energy. Ecol. Indic. 67,
267–282.
Alper, A., Oguz, O., 2016. The role of renewable energy consumption in economic
growth: evidence from asymmetric causality. Renew. Sustain. Energy Rev. 60,
953–959.
Anderson, D., Cavendish, W., 2001. Dynamic simulation and environmental policy
analysis: beyond comparative statistics and the environmental Kuznets curve. Oxf.
Econ. Pap. n 53 (4), 721–746.
Andreoni, J., Levinson, A., 1998. The Simple Analytics of the Environmental Kuznets
Curve. NBER Working Papers 6739.
Andreoni, J., Levinson, A., 2001. The simple analytics of the environmental Kuznets
curve. J. Public Econ. 80, 269–286.
Ang, J.B., 2007. CO2 emissions, energy consumption, and output in France. Energy
Policy 35, 4772–4778.
Anwar, S., Sun, S., 2011. Financial development, foreign investment and economic
growth in Malaysia. J. Asian Econ. 22 (4), 335–342.
Apergis, N., Payne, J.E., 2010a. E energy consumption and economic growth: evidence
from a panel of OECD countries. Energy Policy 38, 656–660.
Apergis, N., Payne, J.E., 2010b. E energy consumption and growth in Eurasia. Energy
Econ. 32, 1392–1397.
Apergis, N., Payne, J.E., 2011. The e energy consumption–growth nexus in Central
America. Appl. Energy 88, 343–347.
Apergis, N., Payne, J.E., 2012b. Renewable and non-renewable energy consumption
growth nexus: evidence from a panel error correction model. Energy Econ. 34,
733–738.
Apergis, N., Payne, J.E., 2014. The causal dynamics between renewable energy, real GDP,
emissions and oil prices: evidence from OECD countries. Appl. Econ. 46 (36),
4519–4525.
Apergis, N., Ozturk, I., 2015. Testing environmental Kuznets curve hypothesis in Asian
countries. Ecol. Indic. 52, 16–22.
Apergis, N., Payne, J.E., 2015. Renewable energy, output, carbon dioxide emissions, and
oil prices: evidence from South America. Energy Sources Part B: Econ. Plan. Policy
10 (3), 281–287.
Balsalobre, D., Álvarez, A.P., 2016. Economic growth and energy regulation in the
environmental Kuznets curve. Environ. Sci. Pollut. Res.. http://dx.doi.org/10.1007/
s11356-016-6773-3.
Balsalobre, D., Álvarez, A., Cantos, J.M., 2015. Public budgets for energy RD & D and the
effects on energy intensity and pollution levels. Environ. Sci. Pollut. Res. 22 (7),
4881–4892.
Balsalobre, D., Álvarez, A., Baños, J., 2016. La innovación y la sustitución energética
como medidas de corrección medioambiental en países de la OCDE. Estud. Econ.
Apl. 34, 235–260.
Ben Jebli, M., Ben Youssef, S., 2015b. Output, e and non-e energy consumption and
international trade: evidence from a panel of 69 countries. Renew. Energy 83,
799–808.
Ben Jebli, M., Ben Youssef, S., Ozturk, I., 2016. Testing environmental Kuznets curve
hypothesis: the role of renewable and non-renewable energy consumption and trade
in OECD countries. Ecol. Indic. 60, 824–831.
Bilgili, F., Koçak, E., Bulut, U., 2016. The dynamic impact of renewable energy
consumption on CO2 emissions: a revisited environmental Kuznets curve approach.
Renew. Sustain. Energy Rev. 54, 838–845.
Bowden, N., Payne, J.E., 2010c. Sectoral analysis of the causal relationship between
renewable and non-renewable energy consumption and real output in the US.
Energy Sources, Part B: Econ., Plan. Policy 5, 400–408.
Bruno, M., Easterly, W., 1998. Inflation Crises and Long-Run Growth. J. Monet. Econ.
41, 3–26.
Bruvoll, A., Fæhn, T., Strøm, B., 2003. Quantifying central hypotheses on environmental
Kuznets curves for a rich economy: a computable general equilibrium study. Scott. J.
Polit. Econ. 50 (2), 149–173.
de Bruyn, Heintz, 1999. The environmental Kuznets curve hypothesis. In: Van den
Bergh, J. (Ed.), Handbook of Environmental and Resource Economics. Edward
Edgar, Cheltenham, UK, 656–677.
de Bruyn, S.M., Van Den Bergh, J.C.J.M., Opschoor, J.B., 1998. Economic Growth and
Emissions. Reconsidering the Empirical Basis of Environmental Kuznets Curves.
Ecol. Econ. 25, 161–175.
Cantos, J.M., Balsalobre, D., 2013. Incidencia del gasto público en I+D+i energético
sobre la corrección medioambiental en España. Estud. Econ. Apl. 31 (1), 93–126.
Cole, M.A., Elliott, R.J.R., Shimamoto, K., 2005. Industrial characteristics,
environmental regulations and air pollution: an analysis of the UK manufacturing
sector. J. Environ. Econ. Manag. 50 (1), 121–143.
Copeland, B.R., Taylor, M.S., 2004. Trade, growth, and the environment. J. Econ. Lit. 42
(1), 7–71.
Costantini, V., Martini, C., 2010. A modified environmental Kuznets curve for sustainable
development assessment using panel data. Int. J. Glob. Environ. Issues 10 (1/2),
84–122.
Costantini, V., Mazzanti, M., 2012. On the green and innovative side of trade
competitiveness? Res. Policy 41, 132–153.
De Leeuw, F., 1962. The demand for capital goods by manufacturers: a study quarterly
A. Alvarez-Herranz et al.
time series. Econometrica 30, 407–423.
Dechezleprêtre, A., 2011. Invention and transfer of climate change mitigation
technologies: a global analysis. Rev. Environ. Econ. Policy 5 (1), 109–130.
Dinda, S., 2004. Environmental Kuznets Curve Hypothesis: a survey. Ecol. Econ. 49 (4),
431–455.
Dogan, E., Seker, F., 2016. Determinants of CO2 emissions in the European Union: the
role of renewable and non-renewable energy. Renew. Energy 94, 429–439.
Dooley, J.J., 1998. Unintended consequences: energy R & D in a deregulated market.
Energy Policy 26 (7), 547–555.
Ewing, B.T., Sari, R., Soyta, U., 2007. Disaggregate energy consumption and industrial
output in the United States. Energy Policy 35, 1274–1281.
Fisher-Vanden, K., Jefferson, G., Liu, H., Tao, Q., 2004. What is driving China's decline
in energy intensity? Resour. Energy Econ. 26, 77–97.
Gangadharan, L., Valenzuela, M.R., 2001. Interrelationships between income, health and
the environment: extending the environmental Kuznets curve hypothesis. Ecol.
Econ. 36 (3), 513–531.
Griliches, Z., 1992. The Search for R & D Spillovers. Scand. J. Econ. 94, 29–47.
Grossman, G., Krueger, E., 1991. Environmental Impacts of a North American Free
Trade Agreement, NBER Working Paper, n. 3914.
Grossman, G., Krueger, E., 1995. Economic growth and the environment. Q. J. Econ. 110
(2), 353–377.
Halicioglu, F., 2009. An econometric study of CO2 emissions, energy consumption,
income and foreign trade in Turkey. Energy Policy 37, 1156–1164.
He, Xiaogang, Zhang, Yaohui, 2012. The influence factors of carbon emissions of Chinese
industry and the reorganization effects of CKC-an empirical research on the dynamic
panel data of different industries based on the STIRPAT model. China Ind. Econ. 1,
26–35.
Heerink, N., Mulatu, A., Bulte, E., 2001. Income Inequality and the Environment:
Aggregation Bias in Environmental Kuznets Curves. Ecol. Econ. 38, 359–367.
Heyes, A., Kapur, S., 2011. Regulatory attitudes and innovation in a model combining
internal and external R & D. J. Environ. Econ. Manag. 61 (3), 327–340.
Kumbaroglu, G.S., 2003. Environmental taxation and economic effects: a computable
general equilibrium analysis for Turkey. J. Policy Model. 25, 795–810.
Lean, H.H., Smyth, R., 2010. CO2 emissions, electricity consumption and output in
ASEAN. Appl. Energy 87, 1858–1864.
Lin, C., Liscow, Z.D., 2013. Endogeneity in the Environmental Kuznets curve: an
instrumental variables approach. Am. J. Agric. Econ. 95, 268–274.
Magnani, N., Vaona, A., 2013. Regional spillover effects of renewable energy generation
in Italy. Energy Policy 56, 663–671.
Markandya, A., Golub, A., Pedrosa-Gallinato, A., 2006. Empirical analysis of national
income and SO2 emissions in selected European countries. Environ. Resour. Econ.
35, 221–257.
Meadows, D.L., Meadows, D.H., Randers, J., Behrens, W., 1972. The Limits to Growth: A
Report for the Club of Rome's Project on the Predicament of Man. Universe Books,
New York.
Mendiluce, M., Pérez-Arriaga, I., OCAÑA, C., 2010. Comparison of the evolution of
energy intensity in Spain and in the EU15. Why Spain Differ.? Energy Policy 38,
639–645.
Menegaki, A.N., 2011. Growth and renewable energy in Europe: a random effect model
with evidence for neutrality hypothesis. Energy Consum. 33, 257–263.
Mensah, T.J., 2014. Carbon emissions, energy consumption and output: a threshold
analysis on the causal dynamics in emerging African economies. Energy Policy 70,
172–182.
Menyah, K., Wolde-Rufael, Y., 2010. CO2 emissions, nuclear energy, renewable energy
and economic growth in the US. Energy Policy 38 (6), 2911–2915.
Moomaw, W., Unruh, G., 1997. Are environmental Kuznets curves misleading us? The
case of CO2 emissions. Environ. Dev. Econ. 2, 451–463.
OECD, 2016. IEA Energy Technology RD & D Statics, 〈http://www.oecd.org/statics〉.
Ozcan, B., 2013. The nexus between carbon emissions, energy consumption and
economic growth in Middle East countries: a panel data analysis. Energy Policy 62,
1138–1147.
Panayotou, T., 1993. Empirical test and policy analysis of environmental degradation at
different stages of economic development (Working Paper 238). Technology and
Environment Programme, International Labour Office, Geneva.
Payne, J.E., 2009. On the dynamics of energy consumption and output in the US. Appl.
Energy 86, 575–577.
397
Energy Policy 105 (2017) 386–397
Payne, J.E., 2010a. On biomass energy consumption and real output in the US. Energy
Sources, Part B: Econ. Plan. Policy 6, 47–52.
Payne, J.E., 2010b. Survey of the international evidence on the causal relationship
between energy consumption and growth. J. Econ. Stud. 37, 53–95.
Payne, J.E., 2010c. A survey of the electricity consumption–growth literature. Appl.
Energy 87, 723–731.
Pizer, W.A., Popp, D., 2008. Endogenizing technological change: matching empirical
evidence to modeling needs. Energy Econ. 30, 2754–2770.
Prieur, F., 2009. The environmental Kuznets curve in a world of irreversibility. Econ.
Theory 40 (1), 57–90.
Romer, P.M., 1990. Are non convexities important for understanding growth?. AEA
Papers and Proceedings 80(2), pp. 97–103.
Sadorsky, P., 2009. E energy consumption and income in emerging economies. Energy
Policy 37, 4021–4028.
Sadorsky, P., 2011. Trade and energy consumption in the Middle East. Energy Econ. 33,
739–749.
Sadorsky, P., 2012. Energy consumption, output and trade in South America. Energy
Econ. 34, 476–488.
Saidi, K., Hammami, S., 2015. The impact of energy consumption and CO2 emissions on
economic growth: fresh evidence from dynamic simultaneous-equations models.
Sustain. Cities Soc. 14, 178–186.
Sari, R., Soytas, U., 2007. The growth of income and energy consumption in six
developing countries. Energy Policy 35, 889–898.
Sebri, M., Ben-Salha, O., 2014. On the causal dynamics between economic growth,
renewable energy consumption, CO2 emissions and trade openness: fresh evidence
from BRICS countries. Renew. Sustain. Energy Rev. 39, 14–23.
Selden, T., Song, D., 1994. Environmental quality and development: is there a Kuznets
Curve for air pollution emissions? J. Environ. Econ. Manag. 27 (2), 147–162.
Sevilla, M., Golf, E., Drih, O.M., 2013. Las energías renovables en España. Estud. Econ.
Apl. 31–1, 35–58.
Shafik, N., Bandyopadhyay, N., 1992. Economic Growth and Environmental Quality:
Time-series and Cross-Country Evidence. World Bank Working Papers 904, pp. 1–6.
Shahbaz, M., Zeshan, M., Afza, T., 2012. Is energy consumption effective to spur
economic growth in Pakistan?. New evidence from bounds test to level relationships
and Granger causality tests. Econ. Model. 29, 2310–2319.
Shi, A., 2003. The impact of population pressure on global carbon dioxide emissions
1975–1996: evidence from pooled cross-country data. Ecol. Econ. 44, 29–42.
Smulders, S., Bretschger, L., 2000. Explaining Environmental Kuznets Curves: How
Pollution Induces Policy and New Technologies (Center Discussion Paper, Vol. 2000-
95). Macroeconomics, Tilburg.
Torras, M., Boyce, J., 1998. Income, inequality, and pollution: a reassessment of the
Environmental Kuznets Curve. Ecol. Econ. 25, 147–160.
Tugcu, C.T., Ozturk, I., Aslan, A., 2012. E and non-e energy consumption and economic
growth relationship revisited evidence from G7 countries. Energy Econ. 34,
1942–1950.
Turner, K., Hanley, N., 2011. Energy efficiency, rebounds and the environmental Kuznets
curve. Energy Econ. 33, 709–720.
UK Climate Change Committee, 2008. Building a Low Carbon Economy: The Uks
Contribution to Tackling Climate Change. UK CCC, London.
Vehmas, J., 2005. Energy-related taxation as an environmental policy tool-the Finnish
experience 1990–2003. Energy Policy 33, 2175–2182.
Velthuijsen, J.W., Worrell, E., 2002. The economics of energy. In: Van den Bergh,
J.C.J.M. (Ed.), Handbook of Environmental and Resource Economics. Edward Elgar,
Cheltenham, UK.
Wackernagel, M., Onisto, L., Bello, P., Callejas, A., López, I., Méndez, J., Suárez, A.,
Suárez, M.A., 1999. National natural capital accounting with the ecological footprint
concept. Ecol. Econ. 29, 375–390.
Wang, Zhaohua, Yin, Fangchao, Zhang, Yixiang, Zhang, Xian, 2012a. An empirical
research on the influencing factors of regional CO2 emission: evidence from Beijing
City, China. Appl. Energy 107, 451–463.
Wang, Zhaohua, Zeng, Hualin, Wei, Yiming, Zhang, Yixian G., 2012b. Regional total
factor energy efficiency: an empirical analysis of industrial sector in China. Appl.
Energy 97, 115–123.
World Bank, 2015. World Development Indicators.
Wu, A.D., Zumbo, B.D., 2008. Understanding and using mediators and moderators. Soc.
Indic. Res. 87, 367–392.