Energy Policy 96 (2016) 131–143

Contents lists available at ScienceDirect

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

Carbon emissions, energy consumption and economic growth: An

aggregate and disaggregate analysis of the Indian economy
Ashfaq Ahmad a, Yuhuan Zhao a,b,n, Muhammad Shahbaz c,d, Sadia Bano a,
Zhonghua Zhang a, Song Wang a, Ya Liu a
a
School of Management and Economics, Beijing Institute of Technology, Beijing 100081, China
b
Center for Energy and Environmental Policy Research, Beijing Institute of Technology, Beijing 100081, China
c
Energy Research Center, Department of Management Sciences, COMSATS Institute of Information Technology, Lahore, Pakistan
d
Montpellier Business School, 2300 Avenue des Moulins, 34080 Montpellier, France

H I G H L I G H T S

Relationships among carbon emissions, energy consumption and economic growth are investigated.

The EKC exists at aggregated and disaggregated levels for India.

All energy resources have positive effects on carbon emissions.

Gas energy consumption is less polluting than other energy sources in India.

art ic l e i nf o a b s t r a c t

Article history: This study investigates the long and short run relationships among carbon emissions, energy con-
Received 26 August 2015 sumption and economic growth in India at the aggregated and disaggregated levels during 1971–2014.
Received in revised form The autoregressive distributed lag model is employed for the cointegration analyses and the vector error
12 April 2016
correction model is applied to determine the direction of causality between variables. Results show that a
Accepted 21 May 2016
Available online 1 June 2016
long run cointegration relationship exists and that the environmental Kuznets curve is validated at the
aggregated and disaggregated levels. Furthermore, energy (total energy, gas, oil, electricity and coal)
Keywords: consumption has a positive relationship with carbon emissions and a feedback effect exists between
Carbon emissions economic growth and carbon emissions. Thus, energy-efficient technologies should be used in domestic
Energy consumption
production to mitigate carbon emissions at the aggregated and disaggregated levels. The present study
Economic growth
provides policy makers with new directions in drafting comprehensive policies with lasting impacts on
EKC
India the economy, energy consumption and environment towards sustainable development.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction economic growth worldwide. Since the mid-twentieth century,

world economies have rapidly grown owing to developments in
Climate change is one of the most debated environmental is- research and technology, international investment, international
sues worldwide. Economic development and globalization have market access, and transportation. The rising trend of economic
contributed to raising the levels of CO2 emissions and other activities in the world in the past few decades has reflected the
greenhouse gases in the atmosphere. In the past few decades, considerable consumption of energy. Energy is the “oxygen” that
climate change has become an issue of discussion among political gives life to economic activities. Developing countries have also
leaders and experts. To date, CO2 emissions have contributed more been consuming energy in large amounts for sustainable economic
than 60% to the increase in the effect of greenhouse gases on the development and better living standards. However, energy con-
atmosphere (Ozturk and Acaravci, 2010). sumption is accompanied by pollutants that contribute to climate
Market liberalization and globalization play important roles in change (Alkhathlan and Javid, 2013).
Energy consumption and CO2 emissions are continuously in-
n
Corresponding author. creasing in different countries, especially developing ones such as
E-mail address: [email protected] (Y. Zhao). India, whose emerging economy and rapid economic development

http://dx.doi.org/10.1016/j.enpol.2016.05.032
0301-4215/& 2016 Elsevier Ltd. All rights reserved.
132 A. Ahmad et al. / Energy Policy 96 (2016) 131–143

has caught the attention of the world. India is the fifth and seventh 2015; Saboori and Sulaiman, 2013; Shahbaz et al., 2015; Wolde-
largest country in terms of energy consumption and energy pro- Rufael, 2010; Ziramba, 2015).
duction, respectively. The growth rate of India has been increasing All above mentioned researches only studied single energy
constructively over the past couple of decades. Thus, the Indian sources for India (e.g., Chandran Govindaraju and Tang, 2013; Ti-
government has adopted several energy policies to generate a wari et al., 2013; Shahbaz et al., 2015), and no one considered all
sustainable energy supply that can meet the country's demand major energy sources simultaneously for analysis. This is a big gap
without any constraints in the future. Inevitably, rapid economic in literature on India and we fill this gap by analyzing the re-
growth in the country also resulted in a large amount of carbon lationships among carbon emissions, energy consumption and
emissions, making India the third largest carbon emitter at a economic growth at aggregated and disaggregated (coal, gas, oil
global level to date. and electricity) level with EKC hypothesis. Aggregate EKC analysis
Masih and Masih (1996) noted that the Indian economy is de- provides information about the relationship among energy con-
pendent on energy, which is a major source of carbon emissions. sumption, economic growth and carbon emissions as a whole. But
The Indian government also has serious concerns about the sus- this information is not clear enough for policy making about future
tainability of its economic growth and environment quality. In its energy demand in terms of which energy source is the better
12thfive-year plan (2012–2017), the Indian government set a tar- substitute for environment protection. For this purpose we have
get of 9% GDP growth. Consequently, energy supply should be conducted the disaggregated EKC analysis for each energy source
increased by 6.5% per year to fulfill this target. In 2009, the Indian separately. Disaggregated analysis examines the effect of each
government also committed to a 20–25% reduction of carbon energy source on carbon emissions and economic growth in de-
emissions by 2020 as compared to 2005 (Boutabba, 2014). Today, tails and provides information about which energy source is the
India faces the daunting task of identifying a solution that could better substitute for future energy demand of India. We can also
help achieve maximum economic growth with minimum CO2 overcome the shortfalls of aggregated EKC analysis by applied
emissions. Therefore, the relationship among CO2 emissions, en- disaggregated EKC analysis which provide us complete informa-
ergy consumption and economic growth in India should be un- tion about the relationships between economic growth and carbon
derstood from both aggregated and disaggregated perspectives for emissions with respect to different energy sources (coal, gas, oil
comprehensive policies on energy, economy and the environment. and electricity) usage separately. So, policy makers can compare
Most of the studies on India examined the effect of a single and select the best energy source to meet the future increasing
energy source on total CO2emissions, thereby failing to provide a demand of energy of India and achieve Indian government's target
clear picture for policy making. By contrast, in this study, all major of minimizing carbon emissions without compromising on eco-
energy sources of India are considered and their respective effects nomic growth at the same time.
on CO2 emissions are examined separately. India meets its overall Compared with previous studies, such as Alkhathlan and Javid
energy demand by exploiting crude oil, petroleum products, coal, (2013), our contributions are as follows: Firstly, we studied the
natural gas, electricity and other energy sources. These energy issue of India, it is a very important country and a fast growing
sources play an important role in meeting the increasing demand economy, and it is also the third largest carbon emitter in the
for energy in India; coal, oil and gas respectively account for 44%, world. This study has great importance for Indian policy maker.
22%, and 7% of the total energy consumption in 2012. India is the Secondly, we considered all major energy sources (coal, gas, oil,
third leading coal consumer, the fourth biggest oil consumer and electricity) of India and checked their impact on their respected
eleventh largest gas consumer in the world. The infrastructure of carbon emissions. Thirdly, we used a longer time series data dur-
electric power has increased significantly, but electricity supply ing 1971–2014 for more reliable and efficient results. Fourthly,
remains insufficient to meet the demands of the country. In fact, some previous studies used only ARDL approach to confirm co-
the growth rate of electricity consumption is lower than the eco- integration among variables. We used both Johansen cointegration
nomic growth rate in the last decade. The electricity consumption test and ARDL bond testing approaches for more reliable results.
per capita of India is the lowest in the world (EIA, 2014). Lastly, some previous studies did not conduct EKC analysis due to
Numerous studies have considered the relationships among high multicollinearity among independent variables. We overcame
CO2 emissions, energy consumption and economic growth at the this problem by using a large sample size of time series data from
aggregated level, resulting in mixed empirical findings (e.g., Alam 1971 to 2014 and conducted EKC analysis between carbon emis-
et al., 2011; Ghosh, 2010; Jayanthakumaran et al., 2012; Vidyarthi, sion and economic growth in each energy consumption model.
2013). Considering the increasing impact of climate change, many The rest of the paper is presented as follows: Section 2 presents
researchers (e.g., Acaravci and Ozturk, 2010; Akbostanci et al., a brief review of the literature, the methodology and data sources
2009; Boutabba, 2014; Managi and Jena, 2008; Pao and Tsai, 2010) employed in this study are described in Section 3, a discussion of
have inspected the empirical relationships among CO2 emissions, the empirical analysis and results is provided in Section 4, and the
energy consumption and economic growth. Their results validated concluding notes as well as policy recommendations are given in
the existence of the environmental Kuznets curve (EKC), which Section 5.
defines the turning point of CO2 emissions with respect to eco-
nomic growth.
This study extends the literature by identifying the long and 2. Literature review
short run relationships that result from the disaggregation of CO2
emissions, energy consumption (oil, gas, coal, and electricity) and Three main groups of researchers have examined the re-
economic growth. Some researchers argued that the relationships lationships among carbon emissions, energy consumption and
among carbon emissions, energy consumption and economic economic growth in the literature (Zhang and Cheng, 2009). The
growth at the aggregated level may not provide a clear picture first group concentrated on analyzing the existence of EKC, i.e.,
because different energy sources have different causality results inverted U-shaped relationship between carbon emissions and
on economic growth and carbon emissions (e.g., Wolde-Rufael, economic growth. Kuznets (1955), the proponent of the Kuznets
2010; Yuan et al., 2008; Ziramba, 2009). Consequently, recent curve, described the relationship between economic growth and
studies have focused on the relationships among carbon emis- income inequality. The Kuznets curve was later transformed into
sions, energy consumption and economic growth at the dis- economic growth and environment quality relationship and called
aggregated level (e.g., Alkhathlan and Javid, 2013; Rahman et al., environmental Kuznets curve. In the beginning, Grossman and
 A. Ahmad et al. / Energy Policy 96 (2016) 131–143 133

Krueger (1991) investigated the relationship between economic energy and domestic production or economic development has
growth and environment pollution using urban areas data of 42 been reported by Asafu-Adjaye (2000) for Asian economies, Paul
countries. They concluded that environmental pollution initially and Bhattacharya (2004) for India, Kahsai et al. (2010) for African
increases but starts to decline once the economy achieves the countries, Ozturk and Salah Uddin (2012) for India, Shahbaz and
threshold of real GDP per capita. Stern et al. (1996) then explored Lean (2012) for Tunisia, Nnaji et al. (2013) for Nigeria, Salahuddin
the linkage between income growth and energy pollutants and and Khan (2013) for Australia, Solarin and Shahbaz (2013) for
validated the EKC hypothesis. Vincent (1997) failed to find an in- Angola and Al-mulali et al. (2014) for Latin America. In such a si-
verted U-shaped relationship between income and energy tuation, policies should encourage energy exploration alongside
pollutants. the adoption of energy-efficient technologies in domestic pro-
Friedl and Getzner (2003) investigated the relationship be- duction expansion. On the one hand, any reduction in energy
tween economic growth and carbon emissions in Austria. Their supply will cause a decline in domestic production and ultimately
empirical exercise obtained inverted-U and N-shaped relation- a decline in economic growth. On the other hand, a decline in
ships between growth and energy pollutants. Galeotti et al. (2006) economic growth will cause a corresponding decrease in energy
studied the relationships between economic growth and en- demand.
vironmental degradation of different energy sources and found Third, the unidirectional causality running from economic
inverted-U shaped relationships for OECD countries. Using data growth to energy consumption is called “conservation hypothesis.”
from 94 countries, Cole (2007) inspected the relationships be- Empirically, many studies provided support to the “conservation
tween income and environment quality and found the validity of hypothesis”, including Kraft and Kraft (1978) and Sari et al. (2008)
U-shaped as well as inverted N-shaped relationships between the for USA, Cheng and Lai (1997) for Taiwan, Ghosh (2002) for India,
variables. Halicioglu (2007) for Turkey, Mehrara (2007) for oil exporting
Furthermore, Coondoo and Dinda (2008) analyzed the con- countries, Dhungel (2008) for Nepal, Kwakwa (2012) for Ghana
firmation of the EKC hypothesis in the presence of intra-country and Ishida (2013) for Japan, among others.
income inequality. Their results validated the EKC hypothesis, Fourth, the “neutrality hypothesis” suggests that no causality
demonstrating that income inequality has a significant effect on exists between economic growth and energy consumption. This
carbon emissions. In the case of India, Managi and Jena (2008) hypothesis has been supported by empirical studies like Cheng
found an inverted-U shaped relationship between growth and (1995) and Payne (2009) for USA, Akinlo (2008) for African
environmental productivity, i.e., EKC hypothesis. Romero-Avila countries and Marques et al. (2014) for Greece. In such a condition,
(2008) used panel data approaches with multiple structural breaks the adoption of energy conservation policies could facilitate eco-
and reported the validity of the EKC hypothesis. Akbostancı et al. nomic growth.
(2009) showed that the empirical relationships between economic The third group of studies in the existing literature has in-
development and CO2 emissions are U-shaped, which is the ne- spected the causal relationships among carbon emissions, energy
gation of the EKC hypothesis in the case of Turkey. Acaravci and consumption and economic growth, including Ang (2008) for
Ozturk (2010) studied the relationships among carbon emissions, Malaysia, Jalil and Mahmud (2009); Zhang and Cheng (2009) for
energy consumption and economic growth in European countries China, Soytas and Sari (2009) for USA, Ocal et al. (2013) for Turkey,
during 1961–2005. Their results indicated the validation of the EKC Chindo et al. (2014) for Nigeria, Kuo et al. (2014) for Hong Kong
hypothesis in Italy and Denmark. and Albiman et al. (2015) for Tanzania. However, they reported
Pao and Tsai (2010) also confirmed the validity of the EKC hy- conflicting empirical findings because of the differences in the
pothesis in India, China, Brazil, Russia, and South Africa. Later on, economic backgrounds of the sample countries, measurements of
many researchers empirically investigated the validity of the EKC variables, specifications, and varied economic, energy and en-
hypothesis in single case studies; however, these generated mixed vironmental policies.
results. For instance, inverted U-shaped relationships between For the Indian economy, many empirical studies have in-
income and CO2 emissions have been reported by Lau et al. (2014) vestigated the validity of the EKC hypothesis, along with the causal
for Malaysia, Gao and Zhang (2014) for Sub-Saharan Africa, Al- relationships among carbon emissions, energy consumption and
Mulali et al. (2015b) for Kenya, Bilgili et al. (2016) for OECD economic growth. For example, Boutabba (2014); Bhattacharyya
countries and Wang et al. (2016) for China. By contrast, Al-Mulali and Ghoshal (2010) supported the EKC hypothesis, whereas Mu-
et al. (2015a); Begum et al. (2015); Ozturk and Al-Mulali, (2015); khopadhyay and Chakraborty (2005); Dietzenbacher and Mukho-
Robalino-López et al. (2015) did not find evidence to support the padhyay (2007) could not validate the EKC hypothesis by applying
EKC hypothesis using data from Vietnam, Malaysia, Cambodia and input–output approaches. Similarly, Ghosh (2010) attempted to
Venezuela, respectively. verify the relationships between economic growth and carbon
The second group of the existing literature investigates the emissions and noted the feedback effect. However, aggregated
relationship between energy use and economic growth and pre- level studies present an unclear picture of the causal relationships
sents four distinct hypotheses. among carbon emissions, energy consumption and economic
First, the unidirectional causality running from energy use to growth in the case of India because different energy sources have
economic growth is called “growth hypothesis,” which asserts that different causalities with carbon emissions and economic growth.
energy performs a key role in promoting economic activity and Later on, some researchers began to study the relationships
reduction in energy supply will reduce economic growth (Ozturk among carbon emissions, energy consumption and economic
and Acaravci, 2010; Shahbaz et al., 2015). For example, growth growth at disaggregated level. Chandran Govindaraju and Tang
changes from a change in energy consumption have been reported (2013) explored the relationship between CO2 emissions and
by Soytas and Sari (2003) for G-7 countries, Altinay and Karagol economic growth by adding coal consumption as a variable in the
(2005) for Turkey, Narayan and Smyth (2008) for OECD countries, emissions function for India and China. Their results indicated that
Ghosh (2010) for India, Odhiambo (2011) for South Africa, Vi- a feedback effect exists between carbon emissions and economic
dyarthi (2013) for India and Iyke (2015) for Nigeria. growth and between coal consumption and carbon emissions of
Second, the so-called “feedback hypothesis” states that eco- India. Tiwari et al. (2013) analyzed the EKC hypothesis by adding
nomic growth is the cause of energy consumption just as energy coal consumption in carbon emissions function. They confirmed
consumption is also a cause of economic growth in the Granger the findings of Chandran Govindaraju and Tang (2013) and vali-
sense. As an example, the interdependent relationship between dated the EKC hypothesis. Shahbaz et al. (2015) examined the
134 A. Ahmad et al. / Energy Policy 96 (2016) 131–143

impact of industrialization and coal consumption on carbon capita, LY denotes the natural log of real GDP per capita, and LY2 is
emissions of India and China. They found U-shaped and inverted the natural log of real GDP per capita square. In addition, α0 is a
U-shaped relationships between the industrial growth and carbon constant, and et is an error term.
emissions of India and China. Their analysis indicated a unidirec- Economic growth has different hypotheses, such as β2 ¼ β3 ¼0
tional causality running from industrial growth and coal con- means a relationship exists at level. If β2 40 and β3 ¼ 0, it shows a
sumption to carbon emissions. monotonically increasing relationship. Here, β2 o0 and β3 ¼ 0 in-
Meanwhile, Abbas and Choudhury (2013) studied the re- dicates a monotonically decreasing relationship, whereas β2 40
lationships between economic growth and the electricity use of and β3 o 0 denotes an inverted U-shaped connection between
the agricultural sectors of Pakistan and India at the aggregated and variables. In addition, β2 o 0 and β3 40 presents a U-shaped
disaggregated levels. They found a bidirectional causative effect connection between variables.
between the electricity consumption of the agricultural sector of The disaggregated analysis describes the respective effects on
India and the country's economic growth; however, unidirectional carbon emissions and economic growth of the consumption of
causality exists from economic growth to electricity use in Paki- different energy sources. We employ separate functions for the
stan. Bildirici and Bakirtas (2014) examined the linkages among consumption of different energy sources. In this study, we con-
gas, oil and coal use of BRICTS (Brazil, Russian, India, China, Turkey sider the main energy types consumed in India, such as coal, oil,
and South Africa) to their GDP, and found long run relationships gas and electricity. These sources are presented by the following
among variables for all these countries. Farhani et al. (2014) ex- equation functions.
plored the relationships among the gas consumption, trade and
LCO2 Coalt = α 0 + β1LECCoal + β2LYt + β3LY2t + et . (2)
economic growth of Tunisia and reported the existence of long run t

relationships among variables. Shahbaz et al. (2014) examined the Eq. (2) shows the relationship among CO2 emissions, coal
long and short run relationships between gas consumption and consumption and economic growth.
economic growth in Pakistan and found a feedback effect between
the two variables. Alkhathlan and Javid (2015) explored the re- LCO2Oilt = α0 + β1LECOil + β2LYt + β3LYt2 + et . (3)
t
lationships among economic growth, oil use in the transportation
sector, and CO2 emissions in Saudi Arabia, and concluded that GDP, Eq. (3) is used to analyze the effect of oil consumption and
oil use and CO2 emissions have a positive relationship with one economic growth on carbon emissions.
another; however, they failed to validate the existence of EKC. Nain LCO2Gast = α0 + β1LECGas + β2LYt + β3LYt2 + et .
t (4)
et al. (2015) explored the linkages among carbon emissions, en-
ergy consumption and economic growth at the aggregated and The effect of gas consumption and economic growth on carbon
sectorial levels, and found that CO2 emissions Granger caused emissions is modeled by Eq. (4).
growth and electricity consumption.
To conclude, most of the studies on the Indian economy ana- LCO2Elct = α0 + β1LECElc + β2LYt + β3LYt2 + et . (5)
t
lyzed the relationships among carbon emissions, energy con-
sumption and economic growth at the aggregated level. However, Eq. (5) represents the link among carbon emissions, electricity
in India, energy demand is met by different types of energy consumption and economic growth.
sources, including coal, gas, oil and electricity. Some researchers In (Eqs. (2)–5) α0 is a constant, LCO2Coalt denotes the natural
have considered the relationships among the consumption of log of the carbon emissions from coal consumption, LECCoalt in-
particular energy resources, carbon emissions and economic dicates the natural log of coal consumption, LCO2Oilt represents
growth for policy making. It's necessary to consider all major en- the natural log of carbon emissions from oil consumption, LECOilt
ergy sources of India in the analysis. This study fills the research refers to the natural log of oil consumption, LCO2Gast denotes the
gap by considering each source of energy and its relationships natural log of carbon emissions with the use of gas energy, LECGast
with related carbon emissions and economic growth. It may help is the natural log of gas consumption, LCO2Elctstands for the nat-
policy makers in choosing the best source of energy and in de- ural log of carbon emissions with the use of electricity, LECElct is
veloping a strategy for the best combination of carbon emissions, the natural log of electricity use, and LYt and LY2t are the natural
energy consumption and economic growth in India. logs of real GDP per capita and real economic growth per capita
square, respectively.

3.1.1. Estimation methodology

3. Methodology and data
The autoregressive distributed lag (ARDL) methodology estab-
lished by Pesaran et al. (2001) is applied to examine the coin-
3.1. Methodology
tegration relationships among carbon emissions, energy con-
sumption and economic growth at the aggregated and dis-
We analyze the relationships among carbon emissions, energy
aggregated levels. Many cointegration testing approaches have
consumption (gas, coal, oil, and electricity) and economic growth
been used in the literature, such as those used by Phillips and
at both the aggregated and disaggregated levels using annual
Hansen (1990); Johansen and Juselius (1990); Engle and Granger
frequency data during 1971–2014. We consider all variables in
(1987). However, these approaches do not estimate the structural
logarithmic form to address the issue of heteroskedasticity and
breaks and the order of integration of variables should be unique.
obtain their growth rate by converting series into their log dif-
However, owing to its flexibility, the ARDL approach is more sui-
ferences. The log-linear quadratic form is employed to analyze the
table for this study compared with the three other approaches.
effects of energy consumption and economic growth on carbon
We can use the ARDL approach when the variables are sta-
emissions. Then, we apply the following model for empirical
tionary at I(1) or I(0) or I(1)/I(0); however, ARDL bound testing is
analysis:
not applicable when the variable is stationary at I(2) (Ouattara,
LCO2t = α 0 + β1LEC + β2LYt + β3LY2t + et , (1) 2004). This methodology provides unbiased results of long run
t
relationships (Narayan, 2005) and is more appropriate for small
where LCO2 represents the natural log of carbon emissions per sample sizes (Haug, 2002). According to Laurenceson and Chai
capita, LEC indicates the natural log of energy consumption per (2003), the ARDL model considers sufficient lag under general to
 A. Ahmad et al. / Energy Policy 96 (2016) 131–143 135

specific approach, and the unrestricted version of the ARDL model computing the ARDL F-statistic for the joint significance of lag-
selects accurate lags for data-generating procedures (Shahbaz and level variables. We derived two hypotheses from Eq. (6) for the
Lean, 2012). The ARDL model has the proper specification to re- long run relationships. The first is null hypothesis, i.e., H0:
solve the problem of endogeneity and residual serial correlation β5 ¼ β6 ¼ β7 ¼ β8 ¼0and the second is the alternative hypothesis,
simultaneously (Pesaran et al., 1998). The unrestricted error cor- i.e., H1: β5‡β6‡β7‡β8‡0.
rection model is presented for long and short run relationships as Next, we compare the ARDL F-statistic with the critical values
follows:
(upper bound and lower bound) generated by Pesaran et al.
p p p
(2001); Narayan (2005). If the ARDL F-statistic lies before the
ΔLCO2t = α 0 + ∑ β1iΔLCO2t − r + ∑ β2iΔLECt − r + ∑ β3iΔLYt − r lower critical bound, then no cointegration exists; if the ARDL
i=1 i=0 i=0
p F-statistic crosses the upper critical bound, then a cointegration is
+ ∑ β4iΔLY2t − r + β5LCO2t − 1 + β6LECt − 1 + β7LYt − 1 present. If the ARDL F-statistic falls between the lower and upper
i=0
critical bounds, then the results are inconclusive (Alkhathlan and
+ β8LY2t − 1 + et . (6) Javid, 2013). If all related variable series are integrated at I(1), then
our assessment regarding cointegration is based on the upper
Before applying ARDL, we need to check the characteristics of
critical bound; otherwise, we follow the lower critical bound if all
the data, that is, whether they fulfill the required assumptions or
not. Accordingly, we first apply different types of unit root tests to the series of related variables are integrated at I(0). After con-
examine the stationary level of all the series because the bound firming the existence of a cointegration between the variables, we
testing approach requires that none of the series is stationary at I move to the second step of examining the long and short run re-
(2). Thus, if any series found stationary at second difference, then lationships, expressed as follows:
the generated results of ARDL F-statistic would be unsound p p p
(Ouattara, 2004). Different unit root tests are available, such as PP LCO2t = α1 + ∑ β1iLCO2t − i + ∑ β2iLEC t − i + ∑ β3iLYt − i
(Phillips-Perron) unit root test by Phillips and Perron (1988), ADF i=1 i=0 i=0
(Augmented Dickey–Fuller) unit root test by Dickey and Fuller p

(1979), and KPSS by Kwiatkowski et al. (1992). Perron (1989) + ∑ β4i LYt2− i + εt ,
i=0 (10)
claimed that ADF, PP, and KPSS do not provide accurate results in
the presence of structural breaks in the variables. Some studies
(e.g., Shahbaz et al., 2013; Raza et al., 2015) in the existing litera- p p p
ture also reported that the results would be biased if structural ΔLCO2t = α2 + ∑ β1iΔLCO2t − i + ∑ β2iΔLEC t − i + ∑ β3iΔLYt − i
breaks are not considered because of the non-rejection of the null i=1 i=0 i=0
hypotheses of the unit root test. Thus, the unit root tests should p

accommodate unknown structural breaks to assess the integrating + ∑ β4iΔLYt2− i + ϕETCt −1 + εt ,

i=0 (11)
properties of the variables, considering that India has im-
plemented several economic, energy and environmental reforms where the ECTt-1 coefficient shows the error correction term cal-
for economic sustainability. To do so, we apply the ZA (Zivot-An- culated from the following:
drews) unit root test presented by Zivot and Andrews (1992) to
identify the unknown structural breaks in the series. Zivot and ECTt −1 = LCO2t −1 − a1 − β2i LECt −1 − β3iLYt −1 − β4i LYt2−1 (12)
Andrews's models are as follows:
p
∆Yt = α 0 + α1DL t+βt+ρyt − 1 + ∑ φi ∆yt − i + εt , 3.1.2. Causality analysis
i=1 (7)
The existence of a cointegration leads us to study the causal
links among the series. To do so, we first employ the vector error
p correction model (VECM) to identify the causal relationship di-
∆Yt = α 0 + γ1DTt+βt+ρyt − 1 + ∑ φi ∆yt − i + εt , rection among the series, namely, carbon emissions, energy con-
i=1 (8) sumption and economic growth. The empirical equations of the
VECM Granger causality are modeled as follows:
p p p p
∆Yt = α 0 + α1DL t + γ1DTt+βt+ρyt − 1 + ∑ φi ∆yt − i + εt , ∆LCo2t = a3 + ∑ β1i ∆LCO2t − i + ∑ β1i ∆LECt − i + ∑ β1i ∆LYt − i
i=1 (9) i=1 i=0 i=0
p
where DL is used as the dummy for the one-time variation in the + ∑ β1iΔLYt2− i + φ1ECTt −1ε3t ,
intercept, and the dummy presents 1 before the time break and i=0 (13)
0 from the time break till the end. DT symbolizes the dummy
variable for the onetime shift in the slope, and the dummy uses DT p p p
¼ t  TB (time break) before the time break date and zero from the ∆LECt = a3 + ∑ β1i ∆LCO2t − i + ∑ β1i ∆LECt − i + ∑ β1i ∆LYt − i
time break till the end. We use these three models in the empirical i=0 i=1 i=0
p
analysis.
+ ∑ β1i ΔLYt2− i + φ1ECTt −1 + ε3t ,
Once all the series are confirmed to be stationary at I(1) or I i=0 (14)
(0) or mixed, and none of the series is stationary at I(2), then we
can analyze the cointegration among the variables by employing p p p
the ARDL bound testing methodology. The bound testing approach ∆LYt = a3 + ∑ β1i ∆LCO2t − i + ∑ β1i ∆LECt − i + ∑ β1i ∆LYt − i
is a two-step procedure. The first step is to select the optimal lags i=0 i=0 i=1
of the variables by applying the Schwarz Bayesian criteria (SBC). p

The second step is running the regression to determine the ex- + ∑ β1iΔLYt2− i + φ1ECTt −1 + ε3t ,
i=0 (15)
istence of long run relationships between the variables by
136 A. Ahmad et al. / Energy Policy 96 (2016) 131–143

p p p Table 1
ΔLYt2 = a3 + ∑ β1i ∆LCO2t − i + ∑ β1i ∆LECt − i + ∑ β1i ∆LYt − i Results of unit root analysis without structural break.
i=0 i=0 i=0
p Variable ADF unit root test PP unit root test Decision
+ ∑ β1i ΔLYt2− i + φ1ECTt −1 + ε3t .
i=1 (16) Level First difference Level First difference

The VECM Granger causality classifies the directions of caus- LY2 4.636 [  5.023*] 9.233 [  5.119*] I(1)
ality into long and short run. The statistical significance of ECTt-1 LY 3.856 [  5.645*] 7.688 [  5.689*] I(1)
LCO2 0.839 [  6.792*] 1.098 [  6.777*] I(1)
(error correction term) with a negative sign confirms that a long
LCO2Coal 0.782 [  5.572*] 0.977 [  5.541*] I(1)
run causality exists among the variables. To determine short run LCO2Elc  1.332 [  5.425*]  1.74 [  5.373*] I(1)
causality, we apply the Wald test on differenced and lagged dif- LCO2Gas  1.328 [  5.521*]  1.35 [  5.521*] I(1)
ferenced terms of the independent variables. Similarly, the sig- LCO2Oil  0.403 [  4.497*]  0.42 [  4.461*] I(1)
nificance of the joint causalities (long and short run causality) is LECCoal 0.626 [  6.414*] 0.923 [  6.427*] I(1)
LEC 1.246 [  6.528*] 1.363 [  6.521*] I(1)
also checked using the Wald test.
LECElc  0.049 [  4.493*]  1.13 [  4.578*] I(1)
LECGas  1.136 [  7.757*]  1.27 [  7.779*] I(1)
3.2. Data sources LECOil 0.053 [  4.981*] 0.056 [  4.905*] I(1)

* shows significance at 1% level.

We use annual time-series data covering the 1971–2014 period
for carbon emissions, aggregated and disaggregated energy con-
sumption and economic growth in the case of India. The data for Table 2
Results of unit root analysis with structural breaks.
the said variables are obtained from different sources. For instance,
CO2 emissions per capita from total energy consumption, CO2 Variable ZA test at level ZA test at first difference Decision
emissions per capita from coal consumption, CO2 emissions per
capita from gas consumption and CO2 emissions per capita from t-statistic Break year t-statistic Break year
oil consumption came from the International Energy Agency from
LY  3.069 (2) 1998  8.226 (1)* 2004 I(1)
1971 to 2013 and one year data of 2014 is extrapolated (IEA, LCO2  3.360 (1) 2001  6.604 (2)* 1979 I(1)
2015).1 The data on coal consumption per capita, gas consumption LCO2Coal  3.250 (2) 2001  6.045 (1)* 2004 I(1)
per capita and oil consumption per capita are acquired from British LCO2Elc  2.544 (1) 1995  7.055 (2)* 2003 I(1)
Petroleum statistics (BP, 2015). The data for energy consumption LCO2Gas  3.478 (1) 1995  7.209 (2)* 1981 I(1)
LCO2Oil  2.409 (1) 2003  5.367(3)** 1990 I(1)
per capita, electricity consumption per capita and CO2 emissions LECCoal  4.177 (3) 1999  7.509 (1)* 1990 I(1)
from electricity and heat production and real GDP per capita are LEC  4.378 (2) 2001  8.050 (2)* 1990 I(1)
obtained from the World Development Indicators CD-ROM (2015). LECElc  3.133 (3) 2000  7.989 (1)* 1085 I(1)
LECGas  4.413 (3) 1986  9.011 (3)* 1981 I(1)
LECOil  4.978 (1) 1995  5.888 (1)* 2001 I(1)

4. Empirical analysis and results discussion Lag order is given in parentheses,* and ** show significance at 1% and 5% levels,
respectively.
Initially, we check the stationary level or the unit root proper-
ties of all the series. We apply PP (Phillips and Perron, 1988) and Table 3
ADF (Dickey and Fuller, 1979) unit root tests, and the test results Results of Johansen cointegrationtest.
are shown in Table 1. Table 1 shows that carbon emissions, energy
Hypothesis Trace test Max. Eigen test Cointegration
consumption and economic growth are non-stationary at both
aggregated and disaggregated levels. After taking their first dif- Energy consumption
ferences, all the variables show the same level of integration. Thus, R r0 33.1084* 21.2573** √
all the series are found to be stationary at I(1). R r1 11.8510 9.2496 
R r2 2.6013 2.6013 
Both PP and ADF unit root tests seem to generate misleading
results because of their low explanatory power when series con- Oil consumption
tains structural breaks. Furthermore, these tests are also unable to R r0 37.6882* 26.9600** √
R r1 10.7282 9.3480 
provide evidence of unknown structural breaks in the series. This R r2 1.3802 1.3802 
issue is resolved using the ZA unit root test, which covers the in-
Gas consumption
formation of single unknown structural breaks in the variables.
R r0 50.2787 28.0476** √
The results presented in Table 2 show that energy consumption R r1 22.2311 15.2061 
(from total energy, oil, gas, coal and electricity), carbon emissions R r2 7.0249 7.0249 
(emissions from total energy, coal, gas, oil and electricity) and Electricity consumption
economic growth have unit root issues in the presence of struc- R r0 44.3069* 29.8755* √
tural breaks. Such structural breaks are the outcomes of the eco- R r1 14.4314 11.2397 
nomic recession and reforms in India during the sample period. R r2 3.1916 3.1916 

For instance, the slowdown in the industrial sector in 2000 af- Coal consumption
fected not only economic growth but also energy demand and R r0 33.5100* 26.1891* √
R r1 7.3208 6.9968 
consequently carbon emissions in the country in 2001. Meanwhile,
R r2 0.3240 0.3240 
the implementation of the Coal Conservation Act affected the coal
consumption as well as the coal carbon emissions in 2001. The * and ** show significance at 1% and 5% levels, respectively.
implementation of the Orissa Electricity Act influenced the
electricity demand and the electricity carbon emissions in 1995.
1
We have filled missing value of one year data 2014 by extrapolated method
The adoption of pricing regimes for the Indian gas sector in 1995
with linear approach. This method also used by Zaman et al. (2015) and Zaman affected the gas carbon emissions. Oil consumption was affected
et al. (2016). by the Gulf war because of the 2003 oil price increase, which also
 A. Ahmad et al. / Energy Policy 96 (2016) 131–143 137

Table 4 The same order of integration of the variables leads us to apply

Results of bound testing cointegration analysis. the Johansen cointegration test to examine the cointegration re-
lationships among the variables. The results are reported in Ta-
Estimated Total Oil energy Gas energy Electricity Coal
model energy energy energy ble 3. We find the presence of a single cointegrating vector via the
trace test and maximum eigenvalue test (max. eigen test). As a
Break year 2001 2003 1995 1995 1999 result, the null hypothesis is rejected. At this point, we may con-
ARDL 6.607*a 9.115*a 8.153*a 7.562*b 8.915*b
clude that cointegration is present among carbon emissions, en-
F-statistic
Optimal lag 1, 2, 1 2, 1, 2 2, 2, 2 2, 1, 1 2, 2, 2
ergy consumption and economic growth at both aggregated and
order disaggregated levels. The results generated using the Johansen
cointegration test are ambiguous because the role of structural
Pesaran et al., 2001 critical values
Constant Constant and trend breaks in the time series data is ignored.
We solve this issue by adding a dummy variable containing
Significance 1% 5% 1% 5%
level information of the break year identified by the ZA unit root test
Lower bounds 4.29 3.23 5.17 4.01 into the unrestricted version of bound testing approach for the
Upper bounds 5.61 4.35 6.36 5.07 cointegration in the five models, i.e., total energy, coal, gas, oil and
Narayan (2005) critical values electricity. This method is suitable for data with small sample si-
Lower bounds 4.06 2.96 zes. Thus, structural breaks in the data are found based on the
Upper bounds 5.25 3.91 results of the ZA structural break unit root test. For this study, we
R2 0.741 0.779 0.912 0.801 0.618
prefer ARDL for cointegration over Johansen cointegration ap-
Adjusted R2 0.491 0.687 0.843 0.669 0.513
DW 2.011 1.896 2.085 2.003 2.04 proach because of the unbiased results of the former. Before ap-
plying the bound testing approach for cointegration, we select the
“a” denotes “unrestricted intercept and trend,” “b” denotes “unrestricted intercept appropriate lag order using SBC, and the results are shown in Ta-
and no trend.”* shows significance at 1% level.
ble 4 (see row 4). The ARDL F-statistic analysis results shown in
Table 4 (see row 3) reveal that the upper critical bounds are less
affected oil carbon emissions. Furthermore, the results in Table 2 than the calculated ARDL F-statistic; these results seem to reject
show that all the variables are integrated at I(1). The ZA test the null hypothesis, i.e., no cointegration. Hence, the empirical
confirms the robustness of the unit root analysis. findings support the acceptance of the alternative hypothesis. In

Table 5
Results of long run and short run analysis.

Sectors/Variables Total energy Coal energy Gas energy Electricity energy Oil energy

Long term elasticities

Constant  3.7988  4.984*  4.882*  1.490*  3.406***
[  1.232] [  3.852] [  3.114] [6.690] [  1.758]
LY 1.370*** 1.820* 1.273* 3.774* 1.271**
[1.887] [6.149] [3.060] [5.152] [2.533]
LY2  0.094**  0.129*  0.092*  0.283*  0.094*
[  2.062] [  6.093] [  3.004] [  5.411] [  2.687]
LEC 0.975* 1.008* 0.569* 0.802* 0.954*
[5.225] [8.566] [5.070] [6.034] [13.415]
Dummy  0.542**  0.041*  0.014**  0.011**  0.045*
[  2.231] [  1.689] [  2.139] [  2.516] [  3.873]

Short term elasticities

Constant 0.011 0.012 0.042 0.025** 0.006
[1.632] [1.504] [1.633] [2.297] [1.397]
ΔLY 0.031 0.018 0.133**  0.158 0.035
[0.236] [0.018] [2.248] [  0.954] [0.352]
ΔLY2  0.52  0.101  0.243**  0.376  0.367
[  0.241] [  0.034] [  2.698] [  0.159] [  0.229]
ΔLEC 0.760* 0.735* 0.600* 0.338 0.813*
[4.171] [5.736] [5.822] [0.674] [8.433]
Dummy 0.001 0.012 0.005  0.005  0.004
[0.150] [0.161] [  0.144] [  0.508] [  0.734]
ECT  0.307**  0.397**  0.343*  0.330***  0.456*
[  2.404] [  2.573] [  4.365] [  1.902] [  3.329]
R2 0.425 0.552 0.582 0.462 0.485
DW test 1.661 1.649 1.872 1.599 1.996
F-statistic 5.341* 9.118* 10.331* 9.396* 16.123*

Diagnostic tests
Normality 0.123 1.379 0.776 0.206 3.151
(0.890) (0.723) (0.678) (0.775) (0.206)
Serial correlation 1.059 1.806 0.576 0.204 0.415
LM (0.357) (0.179) (0.567) (0.125) (0.959)
Heteroskedasticity 0.425 1.353 0.316 0.473 1.065
White (0.827) (0.240) (0.899) (0.793) (0.394)
Ramsey RESET 1.139 0.879 0.461 0.201 0.178
(0.847) (0.375) (0.647) (0.841) (0.859)
CUSUM Stable Stable Stable Stable Stable
CUSUMSQ Stable Stable Stable Stable Stable

T- statistic and P value are given in [] and () respectively. *,** and ***show significance at 1%, 5% and 10% levels, respectively.
138 A. Ahmad et al. / Energy Policy 96 (2016) 131–143

the presence of structural beaks, cointegration is valid among results are robust and reliable. Furthermore, the Durbin-Watson
carbon emissions, energy consumption and economic growth at (DW) test results show the absence of autocorrelation with high R2
both disaggregated and aggregated levels. Thus, the cointegration values for all the models.

Total energy Coal energy

0.0
6.45

6.40
-0.4

6.35

-0.8
6.30
CO2

CO2
-1.2 6.25

6.20
-1.6
6.15

-2.0 6.10
5.2 5.6 6.0 6.4 6.8 7.2 5.2 5.6 6.0 6.4 6.8 7.2

Y Y

Gas energy Electricity energy

4.44
12.60

12.56
4.40
12.52
4.36 12.48

12.44
CO2

4.32
CO2

12.40

4.28 12.36

12.32
4.24
12.28

4.20 12.24
5.2 5.6 6.0 6.4 6.8 7.2 5.2 5.6 6.0 6.4 6.8 7.2

Y Y

Oil energy
4.30

4.28

4.26

4.24
CO2

4.22

4.20

4.18

4.16
5.2 5.6 6.0 6.4 6.8 7.2
Y

Fig. 1. Environmental Kuznets Curves (EKC) in India.

 A. Ahmad et al. / Energy Policy 96 (2016) 131–143 139

Next, the existence of cointegration forces us to examine fur- emissions by 0.957%, 1.008%, 0.569%, 0.802% and 0.954%, respec-
ther the long and short run impacts of economic growth and en- tively. The dummy variables (slowdown in industrial sector, im-
ergy consumption on carbon emissions at the aggregated and plementation of Coal Conservation Act, Orissa Electricity Act, pri-
disaggregated levels. The results are presented in Table 5, which cing regimes for the gas sector, and Gulf war) have negative and
shows that GDP per capita in linear form has a positive effect on significant effects on carbon emissions induced by total energy,
CO2 emissions, but the nonlinear (squared) term of the real GDP coal, gas, electricity and oil.
per capita is negatively related to CO2 emissions. This result re- Results for the short run are shown in Table 5. As can be seen,
veals that the relationship between economic growth and the EKC also exists as we used gas consumption as the source of
CO2emissions assumes an inverted U-shape. Meanwhile, the in- energy. For the other energy sources, the EKC is valid with sta-
verted U-shaped relationship between economic growth and CO2 tistical insignificance. The effects of energy consumption, gas
emissions confirms the existence of EKC. These findings are similar consumption, coal consumption and oil consumption on carbon
with those reported in previous studies (e.g., Bhattacharyya and emissions induced by total energy, gas, coal and oil sources are
Ghoshal (2010); Khanna and Zilberman (2001); Tiwari et al. positive and significant at the aggregated and disaggregated levels.
(2013)) but contradictory to the findings of Chandran Govindaraju However, the effects of the dummy variables are mixed but in-
and Tang (2013); Mukhopadhyay and Chakraborty (2005); Diet- significant. The statistical significance of ECTt-1 also describes the
zenbacher and Mukhopadhyay (2007). The empirical results con- percentage speeds of adjustment from the short run toward the
firm the validity of the EKC hypothesis when oil, gas, coal, and long run equilibrium path. The estimates of ECTt-1 at the ag-
electricity are considered as energy sources. gregated and disaggregated levels are shown in Table 5. As can be
Tiwari et al. (2013) and Shahbaz et al. (2015) also provided seen, the coefficient of ECTt-1 at the aggregated level is  0.307,
disaggregate analysis for India in terms of coal consumption and indicating that the short run deviations in carbon emissions move
our results are consistent with theirs, they also confirmed the toward long run equilibrium path by 30.7% every year.
cointegration and inverted U-shape EKC relationship and showed At the disaggregated level, the ECTt-1 coefficients of carbon
the similar results that increase in coal consumption lead to in- emissions functions of coal, gas, electricity and oil are
crease in carbon emissions. Chandran Govindaraju and Tang  0.397, 0.343,  0.330 and  0.456, respectively. The short run
(2013) studied the relationships among coal consumption, carbon speeds of adjustment of carbon emissions functions of coal, gas,
emissions and economic growth for India but their results are electricity and oil are 39.7%, 34.3%, 33%, and 45.6%, respectively.
contradictory with ours. They concluded that no long-run coin- These results show that the speed of adjustment toward the long
tegration exists among the variables. The reason for this difference run equilibrium of the oil energy model is faster than those of
is that they used whole carbon emissions of India along with coal other energy source models. Such an outcome can be attributed to
consumption and economic growth, while it is carbon emissions several factors, such as an increase in oil prices, less use of oil
from coal consumption used in our study. Khalid Alkhathlann and energy and efficient use of technology. As a consequence of the
Muhammad Javid (2013) provided both the aggregate and dis- increase in oil prices, most consumers and manufacturers in-
aggregate analysis for Saudi Arabia. Their results are also not the troduce efficient technologies to reduce their production costs.
same as ours, they found monotonically increase in economic The speed of adjustment toward the long run equilibrium of
growth and carbon emissions, it's maybe due to countries’ situa- electricity is lower than those of other energy sources. These re-
tions are different, however they showed the similar results about sults reveal that, from the perspective of economic growth and
all energy sources leads to increase in carbon emissions and gas environmental protection, gas energy is better than other energy
energy source is more environment friendly than other energy sources.
resources in Saudi Arabia. We also perform diagnostic tests, and the results are presented
The graphical pictures of the EKCs at the aggregated and dis- in Table 5. The results show the absence of the problem of serial
aggregated levels with the threshold level of the real GDP per correlation, non-normality of the residual terms and hetero-
capita are shown in Fig. 1. The threshold levels of the real GDP per skedasticity, as well as misspecification at the aggregated and
capita for energy consumption per capita, coal consumption per disaggregated levels. Cumulative sum (CUSUM) and cumulative
capita, gas consumption per capita, electricity consumption per sum of squared recursive residual (CUSUMSQ) tests are performed
capita and oil consumption per capita are US$ 1465, US$1157, US to confirm the stability of the long and short run estimates. The
$787, US$1011 and US$863, respectively. These threshold points results are shown in Fig. 2. The figures of CUSUM and CUSUMSQ
describe the differences in the effects of the different energy lie between the upper and lower critical (red lines in the diagram)
sources on CO2 emissions with respect to economic growth. bounds at 5% levels. This result corroborates the reliability and
As mentioned above, a disaggregated EKC analysis helps iden- stability of the long and short run estimates.
tify the least polluting energy source among all the types of energy After analyzing the short run and long run impacts of economic
sources for economic growth. The results show that the use of gas growth and energy consumption on carbon emissions at both
energy results in less CO2 emissions compared with the use of disaggregated and aggregated levels, we then employ the VECM
other energy sources; furthermore, the threshold point for CO2 Granger causality method to determine the causal relationships
emissions is achieved early with respect to economic growth. between variables. The results are reported in Table 6. In the long
Thus, exploring a less polluting alternative energy for economic run, a feedback effect is present between energy consumption and
growth may help expedite the achievement of the threshold point CO2 emissions. Bidirectional causality is also found between en-
of total CO2 emissions. For Indian economic growth to be stable, ergy consumption and growth. Carbon emissions Granger cause
attention should be given to the development and use of better economic growth, and economic growth, in turn, Granger causes
technology. Policy makers should thus plan a strategic combina- carbon emissions.
tion of all energy resources for sustainable economic growth and At the disaggregated level, coal consumption and CO2 emis-
environmental quality. sions have a bidirectional causality. A feedback effect is found
The relationship between energy consumption (gas, coal, oil between coal consumption and CO2 emissions. Economic growth
and electricity) and carbon emissions is positive and statistically Granger causes coal consumption to CO2 emissions. Economic
significant. Keeping other economic agents constant, 1% increases growth and carbon emissions are interdependent, that is, they
in total energy consumption, coal consumption, gas consumption, demonstrate the feedback effect. Gas consumption (economic
electricity consumption and oil consumption increase carbon growth) causes carbon emissions, and CO2 emissions cause gas
140 A. Ahmad et al. / Energy Policy 96 (2016) 131–143

Fig. 2. CUSUM and CUSUMSQ results. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
 A. Ahmad et al. / Energy Policy 96 (2016) 131–143 141

Table 6
Results of Granger causality analysis.

Models Short term χ2statistics Long term t-statistics Short term & long term χ2 Statistics

ΔLCO2 ΔLEC ΔLY ΔLY2 ECT ΔCO2 & ECT ΔEC & ECT ΔY & ECT ΔY2 & ECT

Total energy ΔLCO2  35.179 7.623 7.884  3.24  44.00 23.743 24.767
(0.000) (0.006) (0.005) (0.003) (0.000) (0.000) (0.000)

ΔLEC 21.396  10.329 12.321  4.017 36.438  26.768 29.08

(0.000) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000)

ΔLY 4.702 13.409  1244.49  2.646 8.028 13.418  1518.46

(0.030) (0.000) (0.000) (0.012) (0.018) (0.001) (0.000)

ΔLY2 4.809 18.621 1521.66   2.538 7.803 18.905 1730.95 
(0.028) (0.000) (0.000) (0.016) (0.020) (0.000) (0.000)

Coal energy ΔLCO2  34.998 0.210 0.231  2.939  38.476 10.209 10.199
(0.000) (0.647) (0.631) (0.005) (0.000) (0.006) (0.006)
ΔLEC 0.781 0.051 3333.9   0.914 2.117 0.839 3380.25 
(0.377) (0.000) (0.367) (0.347) (0.000) (0.002)

ΔLY 3.422 0.285  1348.92  2.055 5.494 4.261  1778.52

(.064) (0.593) (0.000) (0.047) (0.064) (0.119) (0.000)

ΔLY2 0.781 0.051 3333.9   0.914 2.117 0.839 3380.25 
(0.377) (0.821) (0.000) (0.367) (0.347) (0.657) (0.000)

Gas energy ΔLCO2  83.721 7.516 6.163  4.389  93.124 24.239 23.852
(0.000) (0.006) (0.013) (0.000) (0.000) (0.000) (0.000)

ΔLEC 87.889  3.517 2.319  9.457 257.74  96.4 95.318

(0.000) (0.061) (0.128) (0.000) (0.000) (0.000) (0.000)

ΔLY 11.063 1.271  4898.88  2.078 21.136 4.918  4965.29
(0.004) (0.259) (0.000) (0.046) (0.000) (0.085) (0.000)
ΔLY2 3.7373 2.3128 3530.16   1.743 6.205 4.749 3600.69 
(0.053) (0.128) (0.000) (0.090) (0.044) (0.093) (0.000)

Electricity energy ΔLCO2  8.714 7.457 6.254  2.394  12.479 11.079 10.239
(0.068) (0.024) (0.044) (0.024) (0.028) (0.011) (0.017)
ΔLY 1.297 0.041  6180.1  2.893 9.795 9.826  6678.96
(0.254) (0.841) (0.000) (0.006) (0.007) (0.007) (0.000)
ΔLY2 0.169 0.103 3334.95   3.334 11.156 11.912 3540.25 
(0.681) (0.739) (0.000) (0.002) (0.003) (0.002) (0.000)

Oil energy ΔLCO2  36.258 1.219 1.447  3.251  40.16 11.008 11.067
(0.000) (0.027) (0.229) (0.003) (0.000) (0.004) (0.004)
ΔLEC 23.198  0.303 0.172  3.342 40.868  11.253 11.201
(0.000) (0.581) (0.679) (0.002) (0.000) (0.003) (0.003)
ΔLY 0.341 0.449  3349.8  2.049 4.326 4.205  3464.26
(0.559) (0.503) (0.000) (0.048) (0.115) (0.122) (0.000)
ΔLY2 0.375 0.253 3310.3   1.805 3.414 3.262 3380.4 
(0.541) (0.615) (0.000) (0.079) (0.181) (0.196) (0.000)

P values are given in parentheses.

consumption (economic growth) in a Granger sense. The re- 5. Conclusions and policy implications
lationship between gas consumption and economic growth is bi-
directional. Similarly, a bidirectional causal relationship exists The main objective of this study is to examine the long and
between electricity consumption and CO2 emissions, between short run relationships among carbon emissions, energy con-
electricity consumption and economic growth, and between eco- sumption (aggregated and disaggregated) and economic growth of
nomic growth and CO2 emissions. The causal relationships among India during 1971–2014 by applying the EKC hypothesis to the
oil consumption, economic growth and CO2 emissions are Indian economy. The cointegration among variables is examined
bidirectional. by applying the ARDL approach. The VECM Granger causality test
In the short run, CO2 emissions, energy consumption and is employed to define the causal relationships among carbon
economic growth at the aggregated level have bidirectional re- emissions, energy consumption (aggregated and disaggregated)
lationships. At the disaggregated level, coal consumption and CO2 and economic growth.
emissions are interdependent, but a neutrality effect exists be- The empirical analysis shows the existence of a cointegration
tween economic growth and CO2 emissions and between coal among carbon emissions per capita, real GDP per capita and en-
consumption and economic growth. Gas consumption and CO2 ergy consumption per capita (aggregated and disaggregated). The
emissions are associated bidirectionally. Hence, economic growth results confirm the existence of an inverted U-shaped relationship
causes CO2 emissions in a Granger sense. Furthermore, a feedback in the long run between carbon emissions and economic growth
effect exists between electricity consumption and carbon emis- for aggregated energy and disaggregated energy (oil, coal, gas and
sions and economic growth Granger causes electricity consump- electricity) models. The EKC hypothesis also exists in the short run
tion and CO2 emissions. The linkage between oil consumption and for the gas energy consumption model; however, EKC is invalid for
CO2 emissions is bidirectional, whereas a unidirectional causality the other energy sources. Furthermore, aggregated energy con-
runs from economic growth to oil consumption and CO2emissions. sumption and disaggregated (oil, gas, coal and electricity) energy
142 A. Ahmad et al. / Energy Policy 96 (2016) 131–143

consumption are related to carbon emissions, that is, carbon Policy 37, 861–867.
emissions increase as energy consumption increases. However, gas Akinlo, A.E., 2008. Energy consumption and economic growth: evidence from 11
Sub-Sahara African countries. Energy Econ. 30, 2391–2400.
is less polluting than other energy sources, and coal is more pol- Al-Mulali, U., Fereidouni, H.G., Lee, J.Y., 2014. Electricity consumption from re-
luting than other energy sources. Electricity consumption results newable and non-renewable sources and economic growth: evidence from
in more pollution than gas energy consumption because 67% of Latin American countries. Renew. Sustain. Energy Rev. 30, 290–298.
Al-Mulali, U., Saboori, B., Ozturk, I., 2015a. Investigating the environmental Kuznets
the electricity in India is produced by coal (World Bank, 2011). The curve hypothesis in Vietnam. Energy Policy 76, 123–131.
VECM results also show feedback effects and causalities in the long Al-Mulali, U., Solarin, S.A., Ozturk, I., 2015b. Investigating the presence of the en-
and short run relationships among carbon emissions, energy vironmental Kuznets curve (EKC) hypothesis in Kenya: an autoregressive dis-
tributed lag (ARDL) approach. Nat. Hazards, 1–19.
consumption and economic growth. Alam, M.J., Begum, I.A., Buysse, J., Rahman, S., Van Huylenbroeck, G., 2011. Dynamic
The Indian economy highly depends on energy consumption modeling of causal relationship between energy consumption, CO2 emissions
because of its status as a fast-growing economy. India needs more and economic growth in India. Renew. Sustain. Energy Rev. 15, 3243–3251.
Albiman, M.M., Suleiman, N.N., Baka, H.O., 2015. The relationship between energy
energy to meet the rising demands of production and sustainable
consumption, CO2 emissions and economic growth in Tanzania. Int. J. Energy
economic development in the country. Thus, policy makers should Sect. Manag. 9, 361–375.
consider disaggregated energy resources to develop a strategy for Alkhathlan, K., Javid, M., 2013. Energy consumption, carbon emissions and eco-
the best combination of energy sources to improve environmental nomic growth in Saudi Arabia: an aggregate and disaggregate analysis. Energy
Policy 62, 1525–1532.
quality without affecting economic growth. A polluted environ- Alkhathlan, K., Javid, M., 2015. Carbon emissions and oil consumption in Saudi
ment adversely affects human health, water and agricultural pro- Arabia. Renew. Sustain. Energy Rev. 48, 105–111.
duction. The Indian government can achieve its goal of maximum Altinay, G., Karagol, E., 2005. Electricity consumption and economic growth: evi-
dence from Turkey. Energy Econ. 27, 849–856.
economic growth with minimum carbon emissions by identifying Ang, J.B., 2008. Economic development, pollutant emissions and energy con-
an alternative for coal energy, which can provide heat to the in- sumption in Malaysia. J. Policy Model. 30, 271–278.
dustrial sectors. According to the analysis, gas energy consumption Asafu-Adjaye, J., 2000. The relationship between energy consumption, energy
prices and economic growth: time series evidence from Asian developing
is favorable in improving environmental quality because gas con- countries. Energy Econ. 22, 615–625.
sumption generates less pollution than the use of other energy Begum, R.A., Sohag, K., Abdullah, S.M.S., Jaafar, M., 2015. CO2 emissions, energy
sources. However, the natural gas reserves of India are limited, consumption, economic and population growth in Malaysia. Renew. Sustain.
Energy Rev. 41, 594–601.
whereas natural coal reserves in India is considerable; thus, coal Bhattacharyya, R., Ghoshal, T., 2010. Economic growth and CO2 emissions. Environ.
will remain the primary source of energy in India for the next 20 Dev. Sustain. 12, 159–177.
years. Therefore, gas can be used in the transportation sector to Bildirici, M.E., Bakirtas, T., 2014. The relationship among oil, natural gas and coal
consumption and economic growth in BRICTS (Brazil, Russian, India, China,
minimize carbon emissions. Policy makers should plan a strategic
Turkey and South Africa) countries. Energy 65, 134–144.
combination of all energy resources of India. They should also Bilgili, F., Koçak, E., Bulut, Ü., 2016. The dynamic impact of renewable energy
promote the infrastructure of the country's industrial sector with consumption on CO2 emissions: a revisited Environmental Kuznets Curve ap-
the use of high-level technologies, coal-bed methane and gas-to- proach. Renew. Sustain. Energy Rev. 54, 838–845.
Boutabba, M.A., 2014. The impact of financial development, income, energy and
liquid technologies, which can be used in energy conversion to trade on carbon emissions: evidence from the Indian economy. Econ. Model.
enhance efficiency. In doing so, India can also utilize its huge coal 40, 33–41.
reserves. BP, 2015. BP statistical review of world energy, 〈http://www.bp.com/en/global/cor
porate/energy-economics/statistical-review-of-world-energy.html〉.
At the same time, India should also build more hydroelectric Chandran Govindaraju, V.G.R., Tang, C.F., 2013. The dynamic links between CO2
power plants to produce electricity because the country has many emissions, economic growth and coal consumption in China and India. Appl.
rivers. Hydroelectricity is less polluting and cheaper than coal- Energy 104, 310–318.
Cheng, B.S., 1995. An investigation of cointegration and causality between energy
fired electricity. The country should also remove all subsidies on consumption and economic growth. J. Energy Dev. 21, 73–84.
energy consumption, especially energy from coal consumption. Cheng, B.S., Lai, T.W., 1997. An investigation of co-integration and causality be-
Therefore, energy consumers and producers should reduce their tween energy consumption and economic activity in Taiwan. Energy Econ. 19,
435–444.
consumption by utilizing energy-efficient technologies that could Chindo, S., Abdulrahim, A., Waziri, S.I., Huong, W.M., Ahmad, A.A., 2014. Energy
contribute to the improvement of environmental quality. Finally, consumption, CO2 emissions and GDP in Nigeria. GeoJournal 80, 315–322.
Indian policy makers should consider attracting foreign invest- Cole, M.A., 2007. Corruption, income and the environment: an empirical analysis.
Ecol. Econ. 62, 637–647.
ment for the development of new and efficient technologies; such
Coondoo, D., Dinda, S., 2008. Carbon dioxide emission and income: A temporal
scheme may help them learn about advanced technology, new analysis of cross-country distributional patterns. Ecol. Econ. 65, 375–385.
ideas, managerial skills and training workers in domestic firms to Dhungel, K.R., 2008. A causal relationship between energy consumption and eco-
nomic growth in Nepal. Asia-Pac. Dev. J. 15, 137–150.
further increase economic growth and decrease carbon emissions.
Dickey, D.A., Fuller, W.A., 1979. Distribution of the estimators for autoregressive
time series with a unit root. J. Am. Stat. Assoc. 74, 427–431.
Dietzenbacher, E., Mukhopadhyay, K., 2007. An empirical examination of the pol-
Acknowledgments lution haven hypothesis for India: towards a green Leontief paradox? Env.
Resour. Econ. 36, 427–449.
EIA, 2014. US Energy Information Administration International energy egency, BP
The authors gratefully acknowledge the financial support from statistical review 〈https://www.eia.gov/beta/international/analysis_includes/
the National Natural Science Foundation of China (Grant no. countries_long/India/india.pdf〉.
Engle, R.F., Granger, C.W., 1987. Co-integration and error correction: representation,
71273026, 71173015, 71473017). Meanwhile, the authors would estimation, and testing. Econometrica 55, 251–276.
like to thank the editor of this journal and the reviewers for their Farhani, S., Shahbaz, M., Arouri, M., Teulon, F., 2014. The role of natural gas con-
detailed and helpful comments. sumption and trade in Tunisia's output. Energy Policy 66, 677–684.
Friedl, B., Getzner, M., 2003. Determinants of CO2 emissions in a small open
economy. Ecol. Econ. 45, 133–148.
Galeotti, M., Lanza, A., Pauli, F., 2006. Reassessing the environmental Kuznets curve
References for CO2 emissions: a robustness exercise. Ecol. Econ. 57, 152–163.
Gao, J., Zhang, L., 2014. Electricity consumption–economic growth–CO2 emissions
nexus in Sub-Saharan Africa: evidence from panel cointegration. Afr. Dev. Rev.
Abbas, F., Choudhury, N., 2013. Electricity consumption-economic growth nexus: 26, 359–371.
An aggregated and disaggregated causality analysis in India and Pakistan. J. Ghosh, S., 2002. Electricity consumption and economic growth in India. Energy
Policy Model. 35, 538–553. Policy 30, 125–129.
Acaravci, A., Ozturk, I., 2010. On the relationship between energy consumption, CO2 Ghosh, S., 2010. Examining carbon emissions economic growth nexus for India: a
emissions and economic growth in Europe. Energy 35, 5412–5420. multivariate cointegration approach. Energy Policy 38, 3008–3014.
Akbostancı, E., Türüt-Aşık, S., Tunç, G.İ., 2009. The relationship between income and Grossman, G.M., Krueger, A.B., 1991. Environmental impacts of a North American
environment in Turkey: is there an environmental Kuznets curve? Energy free trade agreement, National Bureau of Economic Research Working Paper
 A. Ahmad et al. / Energy Policy 96 (2016) 131–143 143

Series No. 3914. Pesaran, M.H., Shin, Y., 1998. An autoregressive distributed-lag modelling approach
Halicioglu, F., 2007. Residential electricity demand dynamics in Turkey. Energy to cointegration analysis. Econom. Soc. Monogr. 31, 371–413.
Econ. 29, 199–210. Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approaches to the analysis
Haug, A.A., 2002. Temporal aggregation and the power of cointegration tests: a of level relationships. J. Appl. Econ. 16, 289–326.
Monte Carlo study. Oxf. Bull. Econ. Stat. 64, 399–412. Phillips, P.C., Hansen, B.E., 1990. Statistical inference in instrumental variables re-
IEA, 2015. CO2 Emissions from Fuel Combustion 2015 Edition, International Energy gression with I (1) processes. Rev. Econ. Stud. 57, 99–125.
Agency, Paris. Phillips, P.C., Perron, P., 1988. Testing for a unit root in time series regression.
Ishida, H., 2013. Causal relationship between fossil fuel consumption and economic Biometrika 75, 335–346.
growth in Japan: a multivariate approach. Int. J. Energy Econ. Policy 3, 127–136. Rahman, M.S., Junsheng, H., Shahari, F., Aslam, M., Masud, M.M., Banna, H., Liya, M.,
Iyke, B.N., 2015. Electricity consumption and economic growth in Nigeria: a revisit 2015. Long-run relationship between sectoral productivity and energy con-
of the energy-growth debate. Energy Econ. 51, 166–176. sumption in Malaysia: an aggregated and disaggregated viewpoint. Energy 86,
Jalil, A., Mahmud, S.F., 2009. Environment Kuznets curve for CO2 emissions: a co- 436–445.
integration analysis for China. Energy Policy 37, 5167–5172. Raza, S.A., Shahbaz, M., Nguyen, D.K., 2015. Energy conservation policies, growth
Jayanthakumaran, K., Verma, R., Liu, Y., 2012. CO2 emissions, energy consumption, and trade performance: evidence of feedback hypothesis in Pakistan. Energy
trade and income: a comparative analysis of China and India. Energy Policy 42, Policy 80, 1–10.
450–460. Robalino-López, A., Mena-Nieto, Á., García-Ramos, J.-E., Golpe, A.A., 2015. Studying
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on the relationship between economic growth, CO2 emissions, and the environ-
cointegration with applications to the demand for money. Oxf. Bull. Econ. Stat. mental Kuznets curve in Venezuela (1980–2025). Renew. Sustain. Energy Rev.
52, 169–210. 41, 602–614.
Kahsai, M., Nondo, C., Schaeffer, P., Gebremedhin, T., 2010. Does level of income Romero-Avila, D., 2008. Questioning the empirical basis of the environmental
matter in the energy consumption and GDP nexus: evidence from Sub-Saharan Kuznets curve for CO2: new evidence from a panel stationarity test robust to
African countries, West Virginia University Regional Research Institute. multiple breaks and cross-dependence. Ecol. Econ. 64, 559–574.
Khanna, M., Zilberman, D., 2001. Adoption of energy efficient technologies and Saboori, B., Sulaiman, J., 2013. Environmental degradation, economic growth and
carbon abatement: The electricity generating sector in India. Energy Econ. 23, energy consumption: evidence of the environmental Kuznets curve in Malaysia.
637–658. Energy Policy 60, 892–905.
Kraft, J., Kraft, A., 1978. Relationship between energy and GNP. J. Energy Dev. 3, Salahuddin, M., Khan, S., 2013. Empirical link between economic growth, energy
401–403. consumption and CO2 emission in Australia. J. Dev. Areas 47, 81–92.
Kuo, K.C., Kanyasathaporn, P., Lai, S., 2014. The causal relationship between GDP, Sari, R., Ewing, B.T., Soytas, U., 2008. The relationship between disaggregate energy
energy consumption and CO2 emissions in Hong Kong. Prod. Res. J. 46, 127–138. consumption and industrial production in the United States: an ARDL approach.
Kuznets, S., 1955. Economic growth and income inequality. Am. Econ. Rev. 45, 1–28. Energy Econ. 30, 2302–2313.
Kwakwa, P.A., 2012. Disaggregated energy consumption and economic growth in Shahbaz, M., Arouri, M., Teulon, F., 2014. Short- and long-run relationships between
Ghana. Int. J. Energy Econ. Policy 2, 34–40. natural gas consumption and economic growth: evidence from Pakistan. Econ.
Kwiatkowski, D., Phillips, P.C., Schmidt, P., Shin, Y., 1992. Testing the null hypothesis
Model. 41, 219–226.
of stationarity against the alternative of a unit root: How sure are we that
Shahbaz, M., Farhani, S., Ozturk, I., 2015. Do coal consumption and industrial de-
economic time series have a unit root? J. Econ. 54, 159–178.
velopment increase environmental degradation in China and India? Environ.
Lau, L.S., Choong, C.K., Eng, Y.K., 2014. Investigation of the environmental Kuznets
Sci. Pollut. Res. 22, 3895–3907.
curve for carbon emissions in Malaysia: do foreign direct investment and trade
Shahbaz, M., Lean, H.H., 2012. The dynamics of electricity consumption and eco-
matter? Energy Policy 68, 490–497.
nomic growth: a revisit study of their causality in Pakistan. Energy 39, 146–153.
Laurenceson, J., Chai, C., 2003. Financial reform and economic development in
Shahbaz, M., Ozturk, I., Afza, T., Ali, A., 2013. Revisiting the environmental Kuznets
China. Edward Elgar Publishing., Cheltenham.
curve in a global economy. Renew. Sustain. Energy Rev. 25, 494–502.
Managi, S., Jena, P.R., 2008. Environmental productivity and Kuznets curve in India.
Solarin, S.A., Shahbaz, M., 2013. Trivariate causality between economic growth,
Ecol. Econ. 65, 432–440.
urbanisation and electricity consumption in Angola: cointegration and caus-
Marques, A.C., Fuinhas, J.A., Menegaki, A.N., 2014. Interactions between electricity
ality analysis. Energy Policy 60, 876–884.
generation sources and economic activity in Greece: a VECM approach. Appl.
Soytas, U., Sari, R., 2003. Energy consumption and GDP: Causality relationship in
Energy 132, 34–46.
G-7 countries and emerging markets. Energy Econ. 25, 33–37.
Masih, A.M., Masih, R., 1996. Energy consumption, real income and temporal
Soytas, U., Sari, R., 2009. Energy consumption, economic growth, and carbon
causality: Results from a multi-country study based on cointegration and error-
emissions: challenges faced by an EU candidate member. Ecol. Econ. 68,
correction modelling techniques. Energy Econ. 18, 165–183.
1667–1675.
Mehrara, M., 2007. Energy consumption and economic growth: the case of oil
Stern, D.I., Common, M.S., Barbier, E.B., 1996. Economic growth and environmental
exporting countries. Energy Policy 35, 2939–2945.
Mukhopadhyay, K., Chakraborty, D., 2005. Environmental impacts of trade in India. degradation: the environmental Kuznets curve and sustainable development.
Int. Trade J. 19, 135–163. World Dev. 24, 1151–1160.
Nain, M.Z., Ahmad, W., Kamaiah, B., 2015. Economic growth, energy consumption Tiwari, A.K., Shahbaz, M., Hye, Q.M.A., 2013. The environmental Kuznets curve and
and CO2 emissions in India: a disaggregated causal analysis. Int. J. Sustain. the role of coal consumption in India: cointegration and causality analysis in an
Energy, 1–18. open economy. Renew. Sustain. Energy Rev. 18, 519–527.
Narayan, P.K., 2005. The saving and investment nexus for China: evidence from Vidyarthi, H., 2013. Energy consumption, carbon emissions and economic growth in
cointegration tests. Appl. Econ. 37, 1979–1990. India. World J. Sci. Technol. Sustain. Dev. 10, 278–287.
Narayan, P.K., Smyth, R., 2008. Energy consumption and real GDP in G7 countries: Vincent, J.R., 1997. Testing for environmental Kuznets curves within a developing
new evidence from panel cointegration with structural breaks. Energy Econ. 30, country. Environ. Dev. Econ. 2, 417–431.
2331–2341. Wang, S., Li, Q., Fang, C., Zhou, C., 2016. The relationship between economic growth,
Nnaji, C.E., Chukwu, J.O., Nnaji, M., 2013. Electricity supply, fossil fuel consumption, energy consumption, and CO2 emissions: empirical evidence from China. Sci.
CO2 emissions and economic growth: implications and policy options for sus- Total Environ. 542, 360–371.
tainable development in Nigeria. Int. J. Energy Econ. Policy 3, 262–271. Wolde-Rufael, Y., 2010. Coal consumption and economic growth revisited. Appl.
Ocal, O., Ozturk, I., Aslan, A., 2013. Coal consumption and economic growth in Energy 87, 160–167.
Turkey. Int. J. Energy Econ. Policy 3, 193–198. World Bank, 2011. Electricity production from coal sources, Retrieved Sept. 9 2011.
Odhiambo, N.M., 2011. Economic growth and carbon emissions in South Africa: an 2015, from 〈http://data.worldbank.org/indicator/EG.ELC.COAL.ZS〉.
empirical investigation. J. Appl. Bus. Res. 28, 37–46. Yuan, J.H., Kang, J.G., Zhao, C.H., Hu, Z.G., 2008. Energy consumption and economic
Ouattara, B., 2004. Foreign aid and fiscal policy in Senegal Mimeo University of growth: Evidence from China at both aggregated and disaggregated levels.
Manchester Manchester. Energy Econ. 30, 3077–3094.
Ozturk, I., Acaravci, A., 2010. CO2 emissions, energy consumption and economic Zaman, K., bin Abdullah, A., Khan, A., bin Mohd Nasir, M.R., Hamzah, T.A.A.T.,
growth in Turkey. Renew. Sustain. Energy Rev. 14, 3220–3225. Hussain, S., 2016. Dynamic linkages among energy consumption, environment,
Ozturk, I., Al-Mulali, U., 2015. Investigating the validity of the environmental Kuz- health and wealth in BRICS countries: green growth key to sustainable devel-
nets curve hypothesis in Cambodia. Ecol. Indic. 57, 324–330. opment. Renew. Sustain. Energy Rev. 56, 1263–1271.
Ozturk, I., Salah Uddin, G., 2012. Causality among carbon emissions energy con- Zaman, K., Ahmad, A., Hamzah, T.A.A.T., Yusoff, M.M., 2015. Environmental factors
sumption growth in India. Èkon. Istraž. 25, 752–775. affecting health indicators in Sub-Saharan African countries: health is wealth.
Pao, H.T., Tsai, C.M., 2010. CO2 emissions, energy consumption and economic Social Indicators Research, pp. 1–14.
growth in BRIC countries. Energy Policy 38, 7850–7860. Zhang, X.P., Cheng, X.M., 2009. Energy consumption, carbon emissions, and eco-
Paul, S., Bhattacharya, R.N., 2004. Causality between energy consumption and nomic growth in China. Ecol. Econ. 68, 2706–2712.
economic growth in India: a note on conflicting results. Energy Econ. 26, Ziramba, E., 2009. Disaggregate energy consumption and industrial production in
977–983. South Africa. Energy Policy 37, 2214–2220.
Payne, J.E., 2009. On the dynamics of energy consumption and output in the US. Ziramba, E., 2015. Causal dynamics between oil consumption and economic growth
Appl. Energy 86, 575–577. in South Africa. Energy Sources Part B Econ. Plan. Policy 10, 250–256.
Perron, P., 1989. The great crash, the oil price shock, and the unit root hypothesis. Zivot, E., Andrews, D., 1992. Further evidence on the great crash, the oil-price shock,
Econometrica 57, 1361–1401. and the unit-root. J. Bus. Econ. Stat. 10, 251–270.