Content uploaded by Sabuhi Yusifov
Author content
All content in this area was uploaded by Sabuhi Yusifov on Dec 27, 2017
Content may be subject to copyright.
114
I
nt. J. Energy Technology and Policy, Vol. 14, No. 1, 2018
Copyright © The Author(s) 2017. Published by Inderscience Publishers Ltd. This is an Open Access Article
distributed under the CC BY license. (http://creativecommons.org/licenses/by/4.0/)
Residential electricity use effects of population in
Kazakhstan
Jeyhun Mikayilov*
King Abdullah Petroleum Studies and Research Center,
P.O. Box 88550, Riyadh 11672, Saudi Arabia
and
Department of Statistics,
Azerbaijan State University of Economics (UNEC),
Istiqlaliyyat Str., 6, Baku, Azerbaijan
and
Institute for Scientific Research on Economic Reforms,
H. Zardabi Avenue, 88a, AZ 1011, Baku, Azerbaijan
Email: jeyhun-mikayilov@unec.edu.az
*Corresponding author
Fakhri Hasanov
King Abdullah Petroleum Studies and Research Center,
P.O. Box 88550, Riyadh 11672, Saudi Arabia
and
Research Program on Forecasting, Economics Department
The George Washington University,
2115 G Street, NW, Washington DC, 20052, USA
and
Department of Socio-Economic Modelling,
Institute of Control Systems,
B. Vahabzade Street 9, Baku, AZ1141, Azerbaijan
Email: fakhri.hasanov@kapsarc.org
Sabuhi Yusifov
Department of Public Administration,
Azerbaijan Technology University,
Shah Ismayil Hatai Ave., 103, Ganja, Azerbaijan
and
The Institute of Economics, ANAS,
Azerbaijan Republic, Baku city,
av. H.Javid, 115, AZ1143, Azerbaijan
Email: s.yusifov@atu.edu.az
Residential electricity use effects of population in Kazakhstan 115
Abstract: We studied impacts of population groups of 15–64 and 65–above on
residential electricity use in Kazakhstan in the STIRPAT framework.
Unlike earlier studies for Kazakhstan in the STIRPAT framework, we applied
time series cointegration and error correction methods. Results from the
autoregressive distributed lags bounds testing approach indicate a significant
impact of the age group of 15–64 on the residential electricity use in long-run,
however, the age group of 65–above has only short-run effects and affluence
has no effect. Another finding is that, 21% of short-run disequilibrium can be
corrected towards long-run equilibrium during a year. Policymakers should
consider the trend of the population group of 15–64 in their decision about the
long-run stance of the residential electricity consumption. The trend suggests
an implementation of energy conservative policy and increasing efficiency of
its usage. Another policy implication is that household’s electricity
consumption is not income dependent maybe due to cheap electricity prices
subsidised by the government. In the short-run, policy makers should consider
the age group of 65–above among other factors in their implementations.
Moreover, they should be careful in making any policy shock to the residential
electricity consumption system, because convergence towards long-run
equilibrium path takes about six years.
Keywords: age groups; residential electricity consumption; STIRPAT;
Kazakhstan; cointegration; error correction modelling; income; Commonwealth
of Independent States; CIS.
Reference to this paper should be made as follows: Mikayilov, J., Hasanov, F.
and Yusifov, S. (2018) ‘Residential electricity use effects of population in
Kazakhstan’, Int. J. Energy Technology and Policy, Vol. 14, No. 1,
pp.114–132.
Biographical notes: Jeyhun Mikayilov is a researcher at the King Abdullah
Petroleum Studies and Research Center in Riyadh, Saudi Arabia. He holds a BS
and MA degrees in Mathematics from Baku State University, Baku/Azerbaijan
and a PhD in Applied Mathematics. His primary research interests include
applied econometrics and sustainable development.
Fakhri Hasanov is a Researcher at King Abdullah Petroleum Studies and
Research Center in Riyadh, Saudi Arabia. His research experience spans
building and applying energy-macroeconometric models for policy purposes,
energy economics with a particular focus on natural resource-rich countries. He
has served as a Deputy Director of the Research Institute at the Ministry of
Economic Development, and a Senior Economist at the Research Department
of the Central Bank, Azerbaijan. He is also affiliated with Department of
Socio-economic Modeling at Institute of System Controls, Azerbaijan. He
holds a PhD in Economics.
Sabuhi Yusifov is currently working at Azerbaijan Technology University
and is Vice Rector for Science and International Relations. He holds a
PhD in Economics. His teaching and research fields are good governance,
decentralisation, local government finance, intergovernmental fiscal relations
and rural development. He was a facilitator and organiser of number of
programs addressed to municipal capacity building in Azerbaijan. These
include local budgeting, municipal entrepreneurship, municipal corporations,
and public-private partnership mechanisms. He is the author of several books
and papers.
116
J
.
M
ikayilov et al.
1 Introduction
Energy is an indispensable part of our society and economy. Understanding the role of
economic growth and population on energy consumption is critical for policymaking and
economic development. This topic has been analysed by a vast number of studies. One of
the prominent questions in this field is relationship between energy use and
socio-economic factors, which has gained special attention with the pioneering study of
Kraft and Kraft (1978). One can find many different studies devoted to the divergent
aspects of this relationship both in national and in cross-national levels. However, little
attention has been paid to the oil-exporting economies of the Commonwealth of
Independent States (CIS), including Kazakhstan. In this regard, as one of the fast growing
economies of the Central Asia, it is valuable to investigate the topic in Kazakhstan.
The results of the research on the relationship between energy use and
socio-economic factors are ambiguous. Therefore, there are four hypotheses in
energy-growth literature (Bozoklu and Yilanci, 2013; Damette and Seghir, 2013; Ozturk,
2010; among others). The growth hypothesis implies that energy consumption is one of
the drivers of economic growth while the conservative hypothesis argues for
unidirectional causality from economic growth to energy consumption. The feedback
hypothesis suggests a bi-directional causal relationship between energy consumption and
economic growth. The last view is neutrality hypothesis, which claims that there is no
causality between energy consumption and economic growth.
As Liddle (2013) expresses, many studies analysing energy use take it as a function
of per capita GDP and price (Holtedahl and Joutz, 2004; Halicioglu, 2007; Dergiades and
Tsoulfidis, 2008; Narayan et al., 2007). However, the impact of population age groups on
energy was not considered by any of the above-mentioned studies. In his work, Liddle
(2013) used panel data for 31 developed countries and 54 developing countries to analyse
the effect of population and its age groups on residential energy consumption. Despite
conducting a large panel analysis, Kazakhstan, the focus of this study, was not included
in his analysis. Meanwhile, numerous valuable studies investigated the relationship of
environmental effects of energy use and age structure of population. Two main directions
draw attention in the energy studies: energy-affluence and energy-population
relationships, which can/need to be combined by the unique model.
The STIRPAT model, developed by Dietz and Rosa (1994, 1997), allows for analysis
of the impacts of population and economic factors on energy use (Liddle, 2014). Almost
all of the energy studies based on the STIRPAT framework investigate the relationship of
the variables of interest using panel or cross-national data for a group of developed and
developing countries (York et al., 2003b; Poumanyvong et al., 2012; York, 2007; Liddle
and Lung, 2010). Regarding Kazakhstan, only a few studies analysed impacts of
economic growth and population or its age groups on energy use by employing STIRPAT
and other modelling frameworks (Brizga et al., 2013). In addition, many of these studies
employed cross-sectional or panel data (Shafiei, 2013; Scarrow, 2010; Fang et al., 2012;
Nouri et al., 2012). By applying STIRPAT framework, Hasanov et al. (2016) examined
the impact of population, age groups and GDP growth on energy use of Azerbaijan,
Kazakhstan and Russia. However, their dependent variable was total energy use and they
did not examine the effects on sectoral energy use, such as industrial or residential
consumption. To our knowledge, this is the first Kazakhstan focused study that examines
the effects of population and affluence on residential electricity consumption (REC)
using the STIRPAT framework, time series cointegration, and error correction modelling.
Residential electricity use effects of population in Kazakhstan 117
This study aims to reveal long- and short-run relationships between the
above-mentioned variables as well as convergence effects for Kazakhstan in the
STIRPAT modelling framework. In the cointegration context, convergence effects, also
known as speed of adjustment (SoA hereafter), provides useful information about the
timeframe needed for the short-run deviation of the relationship to converge towards the
long-run path. Therefore, it is of great importance for policymakers when they develop
measures for managing the growth of electricity use.
We applied the time series cointegration and error correction modelling approach of
the autoregressive distributed lags bound testing (ARDLBT hereafter) to the Kazakhstani
data over the period of 1999–2012. We found a significant impact of the population age
group 15–64 on the residential electricity use in long-run, however, the age group of
65–above has only short-run effects and affluence has no effect. Estimations also
revealed out 21.3% of SoA in the relationship between Kazakhstan’s residential
electricity use and population.
We would expect this study to contribute to the literature as follows: it is the pioneer
study dedicated to residential electricity effects of population and affluence for
Kazakhstan in a time-series analysis. Panel studies have some weaknesses (Kasprzyk
et al., 1989; Dietz and Rosa, 1994; Hsiao, 2003), and country specific features are more
easily discovered in studies using time series analyses as they offer better representation
of country specific features, thus enable more reliable policy recommendations.
Furthermore, as Liddle (2013) expresses, the non-stationarity properties of data were
not considered in earlier STIRPAT-based studies, except for Poumanyvong and Kaneko
(2010) and Liddle (2011). As Liddle (2014) puts it, earlier studies might possibly contain
spurious regression results as economic and population data are commonly
non-stationary.1 By taking non-stationarity of data into account, Liddle (2011, 2013,
2014) and Poumanyvong and Kaneko (2010) applied unit root (UR) and cointegration
tests and estimated long-run elasticities. However, they did not estimate error correction
models and SoA coefficients. The review of existing literature reveals that only Shafiei
(2013) employed panel error correction modelling and estimated SoA coefficient in the
STIRPAT framework. Her study focuses on renewable and non-renewable energy use
effects of population and affluence where she applied panel cointegration and ECM in the
STIRPAT modelling framework for 29 OECD countries. As her focus is on OECD
countries, Kazakhstan is not among the countries studied. Moreover, country specific
aspects are overlooked which, is common for panel analysis, as mentioned before. In the
light of the above-mentioned shortcomings, we accounted for the integration and
cointegration properties of the data and estimated SoA coefficient.
Unlike previous studies on electricity use, we employed Pesaran’s (2001) ARDLBT
approach, to test for cointegration and then to estimate long-run and short-run elasticities
as well as SoA coefficient in the STIRPAT framework. One of the advantages of the
ARDLBT approach is that it works better with small sample sizes and brings about much
more consistent and unbiased estimates (Pesaran et al., 2001; Sulaiman and Muhammad,
2010; Oteng-Abayie and Frimpong, 2006).
The findings of this study may offer useful insights for Kazakhstan’s policymakers so
they can make better electricity market forecasts, and develop adequate measures to
manage the growth of REC. The policymakers should take into account future trend of
the population group aged 15–64 in their decision on the long-run stance of residential
energy consumption. They also should consider that REC in long-run is not income
118
J
.
M
ikayilov et al.
dependent, perhaps due to government subsidisation leading to cheap electricity prices. In
the short-run, the policy makers should consider the population age group of 65–above
among other factors in their REC related measures. Policy makers should also be careful
to avoid creating a policy shock to the REC system, since complete convergence towards
long-run equilibrium path will take about six years.2
The remaining sections of the paper are organised as follows. Section 2 briefly
reviews the existing literature on the STIRPAT analyses of energy use in selected
countries. Section 3 shortly introduces the STIRPAT modelling framework and Section 4
presents the data and describes the econometric method. Section 5 presents and discusses
the results of the empirical analysis. Finally, the main concluding remarks and policy
implications of the study are in Section 6.
2 Brief literature review
A great deal of literature is devoted to studying the energy consumption effects of
population and economic growth by employing STIRPAT modelling framework. Thus, in
this section, we will limit our review only to studies relevant to our research in terms of
methodology used and country chosen.3
As previously noted, some earlier studies have examined this relationship in the
oil-exporting economies of the CIS, including Kazakhstan. However, studies are either
cross-sectional (York et al., 2003a; Knight, 2008; Kick and McKinney, 2014; Lamb
et al., 2014; Mattos and Filippi, 2013), or panel studies (Fang and Miller, 2013;
Martínez-Zarzoso, 2009; York and Rosa, 2012; Brizga et al., 2013; Jorgenson, 2011;
Lankao et al., 2008; Grunewald and Martínez-Zarzoso, 2009a, 2009b, 2011; Prew, 2010;
Iwata and Okada, 2014; Martínez-Zarzoso and Maruotti, 2011) that investigate
environmental issues rather than energy use.4 Only Scarrow (2010), Liddle (2011), Fang
et al. (2012), Nouri et al. (2012), Shafiei (2013), Hasanov et al. (2016) studied the energy
use impacts. Although these energy studies, except Hasanov et al. (2016) and Mikayilov
and Hasanov (2015) are based on panel analysis and ignored country-specific features.
We still review them below.
Shafiei (2013) applied the STIRPAT modelling framework to examine the
determinants of renewable and non-renewable energy consumption for the panel of 29
OECD countries. She chose the period of 1980–2011 and used error correction and
cointegration modelling. Her study found a long-run relationship between the two energy
types and set of the variables: population, its density, GDP per capita, and the GDP share
of service and industry. Coefficients of all regressors were statistically significant in the
long-run elasticity estimations for the non-renewable energy use model. However,
urbanisation and population density were insignificant for the renewable energy use
model. The study results reveal that long-run elasticity with respect to population and
affluence was 1.763 and 0.710 for non-renewable energy, and 0.537, 0.268 for renewable
energy, respectively. Additionally, the estimated SoA coefficients were –0.91 and –0.92
for non-renewable and renewable energy use, which is an indication of rapid convergence
to an equilibrium path.
Scarrow (2010) studied energy consumption effects of population and affluence, with
other explanatory variables, over the period 1960–2007 for a panel of almost all countries
of the world, including Kazakhstan. Results of the employed STIRPAT model showed
that affluence has a statistically significant positive impact on total energy consumption
Residential electricity use effects of population in Kazakhstan 119
and per capita energy consumption. The estimated affluence elasticities of total and per
capita energy consumption varied from 0.03 to 0.2. Total population was found to have a
positive impact (with the coefficients within the interval 0.4–0.8) on total energy use and
negative effect (coefficient was around –0.1) on per capita energy use. However, the
results of the study may suffer from spurious regression problem, because
non-stationarity properties of the data were not taken into account.
Liddle (2011) also used STIRPAT modelling framework to example the
environmental impacts of transport carbon emissions and REC by applying panel FMOLs
to data for 22 OECD countries for the period of 1960–1970. The population variable was
divided into four age groups: 20–34, 35–49, 50–69, and 70 and older. Although
Kazakhstan is not included in this study, it is still relevant to our research since it
conducted a cointegration analysis in the STIRPAT framework. Impacts of population
age structure were statistically significant and different for the age groups. The study
reveals a U-shaped impact of age structure for REC, which means the youngest and
oldest groups have positive effects while middle groups have negative impact. REC
elasticities with respect to population age groups 20–34, 35–49, 50–69 and 70 and above
were 0.219, –0.418, –0.404 and 0.552, respectively.
Nouri et al. (2012) analysed demographic and economic determinants of energy use,
measured in kiloton of oil equivalent for the Economic Cooperation Organization (ECO)
countries, including Kazakhstan over the period of 1960–2012. Panel regression
estimations showed that the driving factors of energy use in the ECO members are total
population, urbanisation and affluence.
Using the STIRPAT framework, Hasanov et al. (2016) examined impacts of total
population, its age groups and affluence on the use of energy in oil-exporting economies
of the CIS: Azerbaijan, Kazakhstan and Russia over the period 1990–2011. An
Autoregressive Distributed Lags Bounds Testing approach was employed in the study.
The study found significant impact of population and its age groups as well as affluence
on the energy use in selected countries. The long-run elasticities of energy use with
respect to the population age group 15–64 in Azerbaijan, Kazakhstan and Russia were
1.92, 0.13 and 8.59, while for the age group of 65 and above the elasticities were 1.71,
0.14, and –1.23, respectively.
Mikayilov and Hasanov (2015) examined impacts of affluence and age groups on
REC in Azerbaijan employing ARDL Bounds Testing approach in the STIRPAT
framework for the period 2000–2012, and concluded that there are significant effects of
population age groups and affluence. The elasticities of REC with respect to population
age group of 15–64 and 65 and above were 10.46 and 2.33, respectively.
In sum, the review of existing literature for the CIS oil-exporting economies shows a
significant gap in this research area. The studies of the economies are mainly panel
studies, which neglect specific features of countries. With the exception of Shafiei
(2013), Hasanov et al. (2016) and Mikayilov and Hasanov (2015), none of these studies
applied cointegration and ECM, and thus have not estimated SoA in the STIRPAT
framework. Finally, except the above mentioned three studies none of them applied the
ARDLBT approach to the time series data of the countries considering that data for these
countries spans for a short period. Hasanov et al. (2016) examined the impact of
population and growth rate on total energy use, but not on residential or industrial energy
use. As a result, their policy suggestions remain too general and do not address specific
energy-type issues. Mikayilov and Hasanov (2015) studied the impacts only in the case of
120
J
.
M
ikayilov et al.
Azerbaijan but not for Kazakhstan. Yet, Shafiei (2013) did not include Kazakhstan in her
analysis.
Saleheen et al. (2012) studied the effect of per capita electricity consumption among
capital, labour and trade openness on economic growth in the production function
framework, applying ARDL BT approach in case of Kazakhstan. They found that 1%
increase in per capita electricity consumption leads to 0.28% increase in per capita GDP.
However, they did not find any casualty from economic growth to electricity
consumption. Moreover, their main interest was economic growth rather than electricity
consumption and therefore, STIRPAT modelling framework have not been applied.
In this study, we consider all of the above-mentioned aspects of the REC in the
STIRPAT framework for Kazakhstan.
3 Framework of analysis: STIRPAT
This section briefly describes the STIRPAT modelling framework employed in our
empirical analysis. STIRPAT is a popular and widely used approach developed by Dietz
and Rosa (1994, 1997). It is based on IPAT, which was first offered by Ehrlich and
Holdren (1971). IPAT assumes that environmental impacts (I) are multiplicative product
of population (P), affluence (A) and technology (T):
I
PAT= (1)
Since IPAT is an accounting identity and therefore assumes proportionality, there is no
space for hypothesis testing. However, the impacts of population, affluence and
technology are certain to differ in magnitude. Primarily because energy use,
environmental, demographic and economic characteristics of the countries differ from
each other. Consequently, IPAT was not featured much in empirical studies. Dietz and
Rosa (1994, 1997) added stochastic terms in equation (1) and thus, the STIRPAT was
developed. The STIRPAT formula can be expressed as the following:
bcd
I
aP A T e= (2)
where a, b, c and d are the coefficients to be econometrically estimated, and e is a
stochastic error term.
Equation (3), which is natural logarithmic expression of equation (2) can be estimated
empirically as:
() * ( ) * ( ) * ( )Ln I q b Ln P c Ln A d Ln T w=+ + + + (3)
where Ln expresses the natural logarithm. q and w are natural logarithm of a and e.
4 Data
In line with the STIRPAT modelling framework, our dataset for Kazakhstan covers the
following indicators:
• REC: The dependent variable in our analysis, measured as the total kilowatt-hours
consumed by residential sector. The data was retrieved from the International Energy
Association (IEA).
Residential electricity use effects of population in Kazakhstan 121
• Gross domestic product (GDP): The sum of gross value added by all resident
producers in the Kazakhstani economy, plus any product taxes, and minus any
subsidies not included in the value of the products (World Bank, 2015). It is
measured in constant 2005 US$.
• Population age group of 15–64 years old (POP_15_64): It is calculated as the share
of the population in the interval of 15–64 years old multiplied by total population,
and measured in persons.
• Population age group of 65 years and older (POP_65): It is calculated as the
population share of 65 years and older multiplied by total population, and measured
in persons.
• Affluence (GDPPC): Measured as GDP per person in constant 2005 US$.
With the exception of REC, the data was retrieved from the World Bank Development
Indicators Database and cover the period of 1999–2012.
Figure 1 illustrates time profile of the variables over the period 1999–2012.
As illustrated in panel A of Figure 1 REC shows an upward trend over the 2000–2012
periods with a level shift in 2005. This level shift may be explained with the economic
developments described below. Strong economic growth observed in Kazakhstan, due to
the record export prices for its energy, minerals, and agricultural goods in 2005.
Government’s efforts to create a good investment environment, especially in the energy
sector, through economic liberalisation and privatisation resulted cumulative 38.4 billion
USD of foreign investment by September 2005. In February 2002, Kazakh Government
established a national energy conglomerate, KazMunaiGaz, combining the national oil
company with the national oil and gas transport company for the stated purpose of being
able to compete with the international oil companies. In 2002, agreement was reached
among Azerbaijan, Georgia, Turkey, Turkmenistan, and Kazakhstan on the route for the
Baku-Tbilisi-Ceyhan (BTC) pipeline, and construction began in May 2003. The first
stage of the pipeline was officially inaugurated in 2005. After re-election in 2005,
President Nazarbaev announced his new economic reforms, which also contributed
significantly to the economic development of the country. Moreover, the construction of
the new energy pipeline between Kazakhstan and China started in 2005.
During the period 2000–2012, population group ages 15–64 and GDP per capita in
Kazakhstan also exhibit increasing trends.
The POP_65 can be characterised by some cyclicality over the period 1990–2012.
The unconventional trend in population age group of 65 and above might be due to the
following reasons. The population began declining in the 1990s due to emigration,
declining fertility rates, and lower life expectancy. Then, the government encouraged
Kazakhs who lived abroad to return. During the ‘90s, there was an organised return
of 70,000 Kazakhs from Mongolia, Iran and Turkey and almost 82,000 Ukrainians,
16,000 Belarussians, 614,000 Russians and 480,000 ethnic Germans returned to
Germany. In addition, many Kazakhs had been displaced internally or had left for other
CIS countries.
122
J
.
M
ikayilov et al.
Figure 1 Time profile of the variables, (a) Panel A: log level of the variables (b) Panel B: growth
rate of the variables (see online version for colours)
4,000,000,000
5,000,000,000
6,000,000,000
7,000,000,000
8,000,000,000
9,000,000,000
10,000,000,000
2000 2002 2004 2006 2008 2010 2012
rec
960,000,000
1,000,000,000
1,040,000,000
1,080,000,000
1,120,000,000
1,160,000,000
2000 2002 2004 2006 2008 2010 2012
pop
_
15
_
64
100,000,000
104,000,000
108,000,000
112,000,000
116,000,000
2000 2002 2004 2006 2008 2010 2012
pop_65
2,000
3,000
4,000
5,000
6,000
2000 2002 2004 2006 2008 2010 2012
gdppc
(a)
-.2
-.1
.0
.1
.2
.3
.4
2000 2002 2004 2006 2008 2010 2012
Diffe
r
enced
r
ec
-.005
.000
.005
.010
.015
.020
.025
.030
2000 2002 2004 2006 2008 2010 2012
Diffe
r
enced pop
_
15
_
64
-.03
-.02
-.01
.00
.01
.02
.03
.04
2000 2002 2004 2006 2008 2010 2012
Differenc ed pop_65
-.04
.00
.04
.08
.12
.16
2000 2002 2004 2006 2008 2010 2012
Differenc ed gdppc
(b)
Note that in the empirical analysis, we used the natural logarithm expressions of the
variables, which are denoted with small letters: rec, gdppc, pop_15_64, pop_65.
Residential electricity use effects of population in Kazakhstan 123
4.1 Econometric method
In the following sub-sections, we discuss the UR tests of Augmented Dickey-Fuller
(ADF) and cointegration methods of the autoregressive distributed lag bounds testing
(ARDLBT).
4.1.1 UR test
It is essential to examine the integration order of variables through UR tests before
conducting a cointegration analysis. To do that, we use the ADF (Dickey and Fuller,
1981) method. Advantages and disadvantages of univariate UR tests, in particular ADF,
have been discussed by Dickey and Fuller (1981), Stock and Watson (1993), Dolado
et al. (1990), Brouwer and Ericsson (1998) and Enders (2010, pp.237–239) among others.
The test takes the null hypothesis of non-stationarity of a given time series.
For a variable y, the ADF statistics are the t-ratio on b1 in the regression below:
011
1
k
ttitit
i
ybψtrend b y y ε
−−
=
Δ= + + + Δ +
∑
α
(4)
Here, Δ and k represent the first difference operator and number of the lags, respectively.
b0 is a constant term, trend and εt are linear time trend and white noise residuals, and i is
lag order. We will skip discussing this test here because of space limitation.
4.1.2 ARDLBT approach
We apply the ARDLBT cointegration approach developed by Pesaran et al. (2001) and
Pesaran and Shin (1999). It is a powerful approach when samples are small, and is easy to
perform by using OLS. Moreover, there is no endogeneity problem in the method, and it
is possible to estimate long and short-run coefficients simultaneously. One of the
advantages of the method is that it can be employed regardless of whether regressors are
I(1), I(0), or a mixture of both (Pesaran et al., 2001; Oteng-Abayie and Frimpong, 2006;
Sulaiman and Muhammad, 2010). This approach is more suitable for our empirical
assessment since the number of observations in our study is relatively small.
Pesaran et al. (2001) describe the following stages of the approach:
a Construction of an unrestricted error correction model (ECM).
01 1
10
nn
ttyxxtitiitit
ii
ycθyθxyφxu
−− − −
==
Δ= + + + Δ + Δ +
∑∑
ϖ
(5)
where y is a depended variable and x is explanatory variable; u indicates white noise
errors; c0 is a drift coefficient; θi denotes long-run coefficients, while i
ϖ
and φi are
short-run coefficients. It is worth mentioning that correct specification of lag length
of the first differenced right-hand side variables in the ARDLBT estimations is one
of the main issues since finding cointegration relationships between variables are
sensitive to lag length [Pesaran et al., (2001), p.23].
Following Pesaran et al. (2001), among others, optimal lag length can be specified
by minimising the Akaike and Schwarz information criteria while removing the
124
J
.
M
ikayilov et al.
serial autocorrelation of residuals. It is advisable to rely on the Schwarz information
criterion when samples are small (Pesaran and Shin, 1999; Fatai et al., 2003).
b After constructing an unrestricted ECM, one can test if cointegrating relationships
exist. The Wald-test (or the F-test) on the θi coefficients above is performed for this
purpose.
The null hypothesis of no cointegration is H0: θ1 = θ2 = θ3 = 0, while an alternative
hypothesis of cointegration is: H1: θ1 ≠ θ2 ≠ θ3 ≠ 0.
If in the given significance level, the computed/sample F-statistic is greater than the
upper bound of the critical value, then one can reject the null hypothesis of no
cointegration. Similarly, the null of no cointegration cannot be rejected if at a given
significance level, the sample F-statistic is smaller than the lower bound of the
critical value. As a third option, the test results will be inconclusive when the sample
value falls between critical values of the upper and low bands.
It is worth mentioning that in the ARDLBT cointegration test, the F-statistics have
non-standard distribution. Thus, F-distribution’s conventional critical values are no
longer valid. Hence, the table of critical values developed by Pesaran and Pesaran
(1997) or Pesaran et al. (2001) must be used.
The cointegrating relationship is stable if θ is statistically significant and negative. In
other words, short-run deviations from the long-run equilibrium path are temporary
and correct towards it.
c If a cointegrating relationship is found among the variables, then the long-run
coefficients can be estimated. We calculate these coefficients based on equation (5)
by either applying a Bewley transformation (Bewley, 1979) or manually setting
c0 + θyt–1 + θyxxt–1 to zero and solving for y as follows.
0yxx
θ
c
yxu
θθ
=− − + (6)
4.2 Small sample bias correction in the ARDLBT approach
Pesaran and Pesaran (1997) used large sample sizes of 500 and 1,000 as well as 20,000
and 40,000 replications, respectively to calculate the F-distribution’s upper and lower
critical values. However, Narayan (2005) mentions that these critical values are
calculated based on large sample points so they are not accurate for small sample sizes
(Narayan, 2004, 2005). Indeed, he compared the critical value generated based on 31
observations with those value reported in Pesaran et al. (2001) at the 5% significance
level and in the case of four regressors. He found that the critical value (3.49) from
Pesaran and Pesaran (1997) is 18.3% lower than his critical value of (4.13). Narayan,
thereby, calculated critical values for small sample sizes ranging from 30 to 80 data
points (see Narayan, 2005). To correct for small sample bias, we employ Narayan’s
critical values in our ARDLBT cointegration test.
Residential electricity use effects of population in Kazakhstan 125
5 Empirical results and discussion
By following the methodological section, we checked integration properties of the
variables using the ADF test. Note that equation (4) is used in all testing exercises. In
other words, a trend and intercept are in all the ADF test specifications regardless of
whether we tested the level or difference of the variables. Our justification is that, as
econometrically explained, if a trend is a part of data generating process and we miss it,
then we will have biased results, which is a serious problem. On the other hand, if a trend
is not a part of data generating process and we have it redundantly, then we will only lose
one degree of freedom.
Table 1 reports the ADF test results.
Table 1 The ADF test results
Variable Panel A: at the level Panel B: at the first
difference Panel C: at the
second difference Conclusion
k Actual value k Actual value k Actual value
rec 0 –2.512278 0 –3.911778** I(1)
gdppc 1 –1.882518 1 –2.917351 1 –4.292989** I(2)
pop_15_64 0 –3.199770 0 –2.073872 1 –5.985103* I(2)
pop_65 2 –5.105226* 2 –5.230702 I(0)
Notes: Maximum lag order is set to two and optimal lag order (k) is selected based on
Schwarz criterion; *, ** and *** indicate statistical significance at the 1%, 5%
and 10% significance levels respectively. The critical values are taken from
MacKinnon (1996). Estimation period: 1999–2012.
As Table 1 concludes, rec is integrated at the order of one, i.e., it is an I(1) process. In
other words, it is non-stationary in its logarithm level, but stationary in its first difference
of logarithm level, which is its growth rate.
On the other hand, the test statistics suggest that gdppc and pop_15_64 are stationary
only after second differencing. Common sense about the integration order of these
variables, however, does not support these statistical results. Moreover, the ADF test
statistics indicate that pop_65 is trend stationary. However, the Phillips-Perron (Phillips
and Perron, 1988) UR test statistics, as well as graphical inspection of the variable
indicates that the series is not trend stationary at level form. We think that such
contrasting results for gdppc and pop_15_64 and pop_65 are mainly caused by the small
number of observations. We have only 14 observations at the best case, which is still
insufficient to get accurate critical values and probabilities. Note that Hasanov et al.
(2016) and Mikayilov and Hasanov (2015) also faced similar problems due to small
sample sizes. In spite of the fact that the test results are quite disappointing, as a research
decision and by relying on the conventional view about integration orders of socio-
economic variables, we consider that all the variables are non-stationary in their log level
and stationary in their growth rate.
We estimated equation (5) for Kazakhstan in two different specifications (i.e., one
with pop_15_64 and another one with pop_65, respectively) as expressed below:5
126
J
.
M
ikayilov et al.
01 12 13 1
10
0
_
15 _ 64
_15_64
tt t t
nn
iti i ti
ii
n
itit
i
rec c θrec θgdppc θpop
ωrec φgdppc
τpop u
−− −
−−
==
−
=
Δ=+ + +
+Δ+ Δ
+Δ +
∑∑
∑
(7)
01 12 13 1
10
0
_
65
_65
tt t t
nn
iti i ti
ii
n
itit
i
rec c θrec θgdppc θpop
ωrec φgdppc
τpop u
−− −
−−
==
−
=
′
′′ ′
Δ=+ + +
′′
+Δ+ Δ
′′
+Δ +
∑∑
∑
(8)
We set the maximum lag order to be one for equations (7) to (8) since the small number
of observations does not allow for the serial correlation LM test to run in more than one
lag order. The test results together with the Schwarz information criterion are tabulated in
Table 2.
Table 2 Statistics for choosing optimal lag size
K SBC FSC(2)
Equation (7) 0* –4.366756 1.987364
[0.2317]
1 –4.563904 31.43241
[0.0308]
Equation (8) 0 –3.859767 0.317201
[0.7381]
1* –4.087104 0.264622
[0.7800]
Notes: k is a lag order while SBC denotes Schwarz information criterion. FSC(2) is the
LM statistics for testing no residual serial correlation against lag orders 1.
Probabilities are in brackets.
We chose zero lag for equation (7) and one lag for equation (8) based on the Schwarz
information criterion and the F-statistics of the serial correlation LM test.
We checked for the existence of long-run (cointegrating) relationships among the
lagged level variables in equations (7) and (8) in the second stage of the ARDLBT
approach. Table 3 reports the cointegration test results.
Two kind of critical values were used in the testing process: Pesaran et al. (2001)
critical values and those from Narayan (2005), with the latter used to avoid potential
biases caused by small sample size of the estimations. Results of the cointegration test
show there is no cointegrating relationship among the lagged level variables in equation
(7). We additionally investigated cointegration properties of REC, population and GDP
per capita. It can be seen from Table 3, that there is a cointegrating relationship at 10%
significance level, between rec and pop_15_64, when we excluded gdppct–1 from the
cointegration space in equation (7). Equation (8) was a similar case, which demonstrated
no long-run co-movement among the lagged level variables of the equation. Furthermore,
detailed investigation of the cointegration relationship among the variables showed that
there is no cointegration relationship even in the combination of any two lagged level
variables in equation (8).
Residential electricity use effects of population in Kazakhstan 127
Table 3 Cointegration test statistics
Equation (7) Equation (8)
Fsample F
sample
3.757330a 1.593500b
Narayan (2005) Pesaran et al. (2001)
At the 1% significance level: 6.760 5.580
At the 5% significance level: 4.663 4.160
At the 10% significance level: 3.797 3.510
At the 1% significance level: 6.265 5.000
At the 5% significance level: 4.428 3.870
At the 10% significance level: 3.695 3.350
Notes: aIn the case of one regressor, restricted intercept and no trend.
bIn the case of two regressors, restricted intercept and no trend.
The first three upper bound critical values of Narayan (2005) and Pesaran et al.
(2001) are in the case of two regressors, restricted intercept and no trend, while
the last three upper bound critical values are in the case of one regressor, restricted
intercept and no trend.
Thus, we found cointegrating relationship between rec and pop_15_64 in equation (7)
and only short-run relationship for the first differenced regressors in the equation (8).
As a next stage of the ARDLBT approach, we specified the final ECM specifications.
The final specifications are presented in Table 4.
Table 4 The final ARDL specification
Regressor Coef. (std. error)
Panel A: the estimated final ARDL specification of equation (7)
rect–1 –0.212997 (0.070586)
pop_15_64t–1 0.972618(0.295416)
Intercept –15.35966 (4.777056)
Δpop_15_64t 0.931658 (1.154106)
Δgdppct 0.323134 (0.196360)
DP05 0.261101 (0.022297)
DP00 –0.164320 (0.022519)
Panel B: the estimated final ARDL specification of equation (8)
Δpop_65t 0.562013 (0.287682)
Δrect–1 –0.158072 (0.063311)
Intercept 0.039739 (0.006894)
DP05 0.273428 (0.022991)
DP00 –0.171509 (0.022916)
Note: Dependent variable is Δrect; method: least squares; estimation period: 1999–2012.
Each of the final specifications in Table 4 succeeded residual diagnostics tests of the
serial correlation, autocorrelation, heteroscedasticity, normality, and Ramsey reset’s
128
J
.
M
ikayilov et al.
misspecification test. We do not report the results of tests here. However, they can be
obtained upon request.
According to the results, the population group ages 15–64 is statistically significant in
the final specification of equation (7). The elasticity of the REC with respect to this age
group is 4.568 (0.973/0.213). Putting it differently, a 1% increase in the age group leads
to a 4.56% increase in REC in the long-run. Despite the higher magnitude of the
coefficient, the sign of it is consistent with the conventional interpretation of the
STIRPAT framework. Note that the growth rates of the age group and affluence are
positive in the final specification of equation (7), which is consistent with the STIRPAT
framework. However, they are statistically insignificant most probably due to small
number of observation.
As panel A of Table 4 demonstrates, SoA coefficient is 0.213, which means 21.3% of
disequilibrium in the short-run can be corrected towards long-run equilibrium path during
a year.
Panel B of Table 4 reports that the growth rate of the population group ages 65 and
above has a positive impact on the growth rate of REC in the short-run. Numerically, a
1% increase in the growth rate of this population group causes a 0.56% increase in the
growth rate of electricity consumption. The positive impact of the population age group
on the electricity consumption is also consistent with the STIRPAT framework.
Moreover, there is a dynamic relationship in REC as its one year lagged growth rate has a
statistically significant impact on the current year’s growth rate.
Additionally, as reported in the both panels of Table 4, dummy variables for
capturing extraordinary decrease and increase in REC in 2000 and 2005 respectively are
statistically significant.
6 Concluding remarks
A great deal of studies has examined energy consumption effects of population, its age
groups, and affluence in developed and developing countries. However, the number of
studies focusing on oil-exporting countries of the CIS, in particular for Kazakhstan, is
limited to either panel or cross sectional analyses. In the absence of time series studies, it
is difficult to discover country specific features of energy effects of population and
affluence. To our knowledge, there is one time series analysis for Kazakhstan, Hasanov
et al. (2016) in the STIRPAT framework, where the dependent variable is aggregated
energy use, which does not allow using specific energy types and offer related policy
suggestions. With these considerations in mind, we examined the impacts of population
age groups and affluence on REC in Kazakhstan by applying an ARDLBT cointegration
method, a powerful method in the case of small sample in the STIRPAT framework. As
the number of observations is small, results of our empirical estimations and conclusions
are presented cautiously. We found that one of the driving forces of the REC in the long-
run is the population age group of 15–64, whereas the age group of 65 and above exhibit
only short-run effects. The affluence was not found to have any statistically significant
influence on the REC. Moreover, the analysis revealed that 21.3% speed of convergence
towards long-run equilibrium path in the relationship of REC and the 15–64 population
age group.
Findings of this study may offer useful insights for policymakers in making better
electricity demand forecasts and taking adequate measures on residential electricity use in
Residential electricity use effects of population in Kazakhstan 129
Kazakhstan. The policymakers should take into account trend of the population group
aged 15–64 in their decision on the long-run stance of residential energy consumption. As
Figure 1 illustrates population age group 15–64, which has a significant long-run effect
on REC, has a strict upward trend over the period of analysis. Such a strong upward trend
in the group is related to national and traditional customs of the Kazakhstan, and
therefore may not easily be curbed in future. Further, the SoA coefficient is as small as
21.3%. Combining these two findings, we can conclude that, in the long-run, adequate
policy measures should include increase in the efficiency of electricity consumption and
applying energy conservation measures. As Liddle (2011) notes, the size of the family is
proportional with the age of family head in general, which is also the case for
Kazakhstan. Promoting more efficient (and economically viable) electricity appliances
with these household is a suggestion for increasing efficiency of electricity usage.
Policymakers should also pay attention to the finding that REC is not income dependent,
perhaps due to cheap electricity prices subsidised by the government. In the short-run, the
policymakers should consider the population age group of 65–above among other factors
in their REC related measures. Furthermore, policymakers should be careful not to create
a policy shock to the REC system, since convergence towards the long-run equilibrium
path takes about six years.
Acknowledgements
The authors are deeply thank to the editor of this journal and anonymous referees for
their comments and suggestions. Proofreading of Patrick Bean and Kenneth White are
greatly appreciated. All remaining errors and omissions are our sole responsibility. Please
note that views expressed in this paper are those of the authors and do not necessarily
represent the views of their affiliated institutions.
References
Bewley, R.A. (1979) ‘The direct estimation of the equilibrium response in a linear model’,
Economics Letters, Vol. 3, No. 4, pp.357–361.
Bozoklu, S. and Yilanci, V. (2013) ‘Energy consumption and economic growth for selected OECD
countries: further evidence from the Granger causality test in the frequency domain’, Energy
Policy, Vol. 63, No. 100, pp.877–881.
Brizga J., Feng K. and Hubacek, K. (2013) ‘Drivers of CO2 emissions in the former Soviet Union:
A country level IPAT analysis from 1990 to 2010’, Energy, Vol. 59, pp.743–753.
Damette, O. and Seghir, M. (2013) ‘Energy as a driver of growth in oil exporting countries?’,
Energy Economics, Vol. 37, No. 100, pp.193–199.
Brouwer, G. and Ericsson, N.R. (1998) ‘Modeling inflation in Australia’, Journal of Business and
Economic Statistics, Vol. 16, No. 4, pp.433–449.
Dergiades, T. and Tsoulfidis, L. (2008) ‘Estimating residential demand for electricity in the United
States.1965–2006’, Energy Economics, Vol. 30, pp.2722–2730.
Desai, P. (2000) ‘Why did the Ruble collapse in August 1998?’, The American Economic Review,
Vol. 90, No. 2, pp.48–52, JSTOR 117190, subscription required.
Dickey, D. and Fuller, W. (1981) ‘Likelihood ratio statistics for autoregressive time series with a
unit root’, Econometrica, Vol. 49, No. 4, pp.1057–1072.
130
J
.
M
ikayilov et al.
Dietz, T. and Rosa, E.A. (1994) ‘Rethinking the environmental impacts of population, affluence,
and technology’, Human Ecology Review, Vol. 1, pp.277–300.
Dietz, T. and Rosa, E.A. (1997) ‘Effects of population and affluence on CO2 emissions’,
Proceedings of the National Academy of Sciences, Vol. 94, pp.175–179, USA.
Dolado, J., Jenkinson, T. and Sosvilla-Rivero, S. (1990) ‘Cointegration and unit roots’, Journal of
Economic Surveys, Vol. 4, No. 3, pp.249–273.
Enders, W. (2010) Applied Econometrics Time Series, 3rd ed., Wiley Series in Probability and
Statistics, University of Alabama, John Wiley and Sons, New York.
Engle, R.F. and Granger, C.W.J. (1987) ‘Co-integration and error correction: representation,
estimation and testing’, Econometrica, Vol. 55, No. 2, pp.251–276.
Ehrlich, P.R. and Holdren, J.P. (1971) ‘Impact of population growth’, Science, Vol. 171, No. 3977,
pp.1212–1217.
Fang, W. and Miller, S. (2013) ‘The effect of ESCOs on carbon dioxide emissions’, Applied
Economics, Vol. 45, No. 34, pp.4796–4804.
Fang, W., Miller, S. and Yeh, C. (2012) ‘The effect of ESCOs on energy use’, Energy Policy,
December, Vol. 51, pp.558–568.
Fatai, K., Oxley, L. and Scrimgeour, F.G. (2003) ‘Modeling and forecasting the demand for
electricity in New Zealand: a comparison of alternative approaches’, The Energy Journal,
Vol. 24, No. 1, pp.75–102.
Grunewald, N. and Martínez-Zarzoso, I. (2009a) ‘Carbon dioxide emissions, economic growth and
the impact of the Kyoto protocol’ [online] http://www.webmeets.com/files/papers/EAERE/
2011/1163/Grunewald%20and%20Martinez-Zarzoso_Carbon%20Dioxide%20Emissions,%
20Economic%20Growth%20and%20the.pdf (accessed 10 February 2016).
Grunewald, N. and Martínez-Zarzoso, I. (2009b) Driving Factors of Carbon Dioxide Emissions
and the Impact from Kyoto Protocol, CESIFO Working paper No. 2758.
Grunewald, N. and Martínez-Zarzoso, I. (2011) How Well Did the Kyoto Protocol Work? A
Dynamic-GMM Approach with External Instruments, Ibero-America Institute for Economic
Research (IAI) Discussion papers 212.
Halicioglu, F. (2007) ‘Residential electricity demand dynamics in Turkey’, Energy Economics,
Vol. 29, No. 2, pp.199–210.
Hasanov, F., Bulut, C. and Suleymanov, E. (2016) ‘Do population age groups matter in the energy
use of the oil-exporting countries?’, Economic Modelling, Vol. 54, pp.82–99.
Holtedahl, P. and Joutz, F.L. (2004) ‘Residential electricity demand in Taiwan’, Energy
Economics, Vol. 26, No. 2, pp.201–224.
Hsiao, C. (2003) Analysis of Panel Data, 2nd ed., Cambridge University Press, UK.
Iwata, H. and Okada, K. (2014) ‘Greenhouse gas emissions and the role of the Kyoto protocol’,
Environmental Economics and Policy Studies, Vol. 16, pp.325–342.
Jorgenson, A.K. (2011) ‘Carbon dioxide emissions in central and eastern European nations, 1992–
2005: a test of ecologically unequal exchange theory’, Human Ecology Review, Vol. 18, No. 2,
pp.105–114.
Kasprzyk, D., Duncan, G., Kalton, G. and Singh, M.P. (1989) Panel Surveys, John Wiley,
New York.
Kick, E.L. and McKinney, L.A. (2014) ‘Global context, national interdependencies, and the
ecological footprint: a structural equation analysis’, Sociological Perspectives, Vol. 57, No. 2,
pp.256–279.
Knight, K.W. (2008) The Ecological Implications of population Aging: A Cross-National Analysis
of the Ecological Footprint, December, Washington State University.
Kraft, J. and Kraft, A. (1978) ‘Note and comments: on the relationship between energy and GNP’,
The Journal of Energy and Development, Vol. 3, pp.401–403.
Residential electricity use effects of population in Kazakhstan 131
Lamb, W.F., Steinberger, J.K., Bows-Larkin, A., Peters, G.P., Roberts, J.T. and Wood, F.R. (2014)
‘Transitions in pathways of human development and carbon emissions’, Environ. Res. Lett.,
Vol. 9, No. 1, p.014011, 10pp.
Lankao, P.R., Nychka, D. and Tribbia, J.L. (2008) ‘Development and greenhouse gas emissions
deviate from the ‘modernization’ theory and ‘convergence’ hypothesis’, Climate Research,
Vol. 38, pp.17–29.
Liddle, B. (2011) ‘Consumption-driven environmental impact and age structure change in OECD
countries: a cointegration-STIRPAT analysis’, Demographic Research, Vol. 24, pp.749–770.
Liddle, B. (2013) ‘Population, affluence, and environmental impact across development: evidence
from panel cointegration modeling’, Environmental Modelling and Software, Vol. 40,
pp.255–266.
Liddle, B. (2014) ‘Impact of population, age structure, and urbanization on carbon emissions/
energy consumption: evidence from macro-level, cross-country analyses’, Population and
Environment, Vol. 35, No. 3, pp.286–304.
Liddle, B. and Lung, S. (2010) ‘Age structure, urbanization, and climate change in developed
countries: revisiting STIRPAT for disaggregated population and consumption-related
environmental impacts’, Population and Environment, Vol. 31, No. 5, pp.317–343.
MacKinnon, J. (1996) ‘Numerical distribution functions for unit root and cointegration tests’,
Journal of Applied Econometrics, Vol. 11, No. 6, pp.601–618.
Martínez-Zarzoso, I. (2009) A General Framework for Estimating Global CO2 Emissions,
Ibero-America Institute for Economic Research, Discussion paper No. 180.
Martínez-Zarzoso, I. and Maruotti, A. (2011) ‘The impact of urbanization on CO2 emissions:
evidence from developing countries’, Ecological Economics, Vol. 70, pp.1344–1353 [online]
http://www.sciencedirect.com/science/article/pii/S0921800911000814 (accessed 23 October
2015).
Mattos, E.J. and Filippi, E.E. (2014) ‘Drivers of environmental impact: a proposal for nonlinear
scenario designing’, Environmental Modelling and Software, Vol. 62, No. C, pp.22–32.
Mikayilov, C. and Hasanov, F. (2015) Residential Electricity Use Effects of Population in
Azerbaijan, Unpublished paper.
Narayan, P.K. (2004) An Econometric Model of Tourism Demand And A Computable General
Equilibrium Analysis of the Impact of Tourism: The Case of the Fiji Islands, Unpublished PhD
thesis, Department of Economics, Monash University, Melbourne, Australia.
Narayan, P.K. (2005) ‘The saving and investment nexus for China: evidence from cointegration
tests’, Applied Economics, Vol. 37, No. 17, pp.1979–1990.
Narayan, V., Ramaswamy, S. and Menon, N. (2007) ‘Long-lived giant number fluctuations in a
swarming granular nematic’, Science, Vol. 317, No. 5834, pp.105–108.
Nouri, M., Mohaghegh, M. and Azizi, A. (2012) A Comparative Study on the Relationship between
Energy Consumption and Main Demographic and Economic Indicators among ECO Member
States, Formal Report, Population studies and Research Center for Asia and the Pacific
[online] http://www.ecosn.org/documents/projects.aspx (accessed 15 August 2015).
Oteng-Abayie, E.F. and Frimpong, J.M. (2006) ‘Bounds testing approach to cointegration: an
examination of foreign direct investment trade and growth relationships’, American Journal of
Applied Sciences, Vol. 3, No. 11, pp.2079–2085.
Ozturk, I. (2010) ‘A literature survey on energy-growth nexus’, Energy Policy, Vol. 38, No. 1,
pp.340–349.
Pesaran, H.M. and Pesaran, B. (1997) Working with Microfit 4.0: Interactive Econometric Analysis,
Oxford University Press, Oxford, UK.
Pesaran, M. and Shin, Y. (1999) ‘An autoregressive distributed lag modeling approach to
cointegration analysis’, in Strom, S. (Ed.): Econometrics and Economic Theory in the 20th
Century: The Ragnar Frisch centennial Symposium, Cambridge University Press, Cambridge.
132
J
.
M
ikayilov et al.
Pesaran, M.H., Shin, Y. and Smith, R.J. (2001) ‘Bounds testing approaches to the analysis of level
relationships’, Journal of Applied Econometrics, Vol. 16, No. 3, pp.289–326.
Phillips, P.C.B. and Perron, P. (1988) ‘Testing for a unit root in time series regression’, Biometrika,
Vol. 75, No. 2, pp.335–346, DOI: 10.1093/biomet/75.2.335.
Poumanyvong, P., Kaneko, S. and Dhakal, S. (2012) ‘Impacts of urbanization on national transport
and road energy use: evidence from low, middle and high income countries’, Energy Policy,
Vol. 46, pp.268–277, DOI: 10.1016/j.enpol.2012.03.059.
Prew, P. (2010) ‘World-economy centrality and carbon dioxide emissions: a new look at the
position in the capitalist world-system and environmental pollution’, Journal of
World-Systems Research, Vol. 16, pp.162–191.
Saleheen, K., Ahmed, J.F. and Muhammad, S. (2012) Electricity Consumption and Economic
Growth in Kazakhstan: Fresh Evidence from a Multivariate Framework Analysis, MPRA
paper No. 43460, posted 28 December 2012 07:24 UTC.
Scarrow, R.M. (2010) Uncovering the Energy Efficiency of the Post-Industrial World: An Analysis
of Ecological Factors in Energy Use Across Nations, 1960–2007, The Ohio State University.
Shafiei, S. (2013) Economic Growth, Energy Consumption, and Environment: Assessing Evidence
from OECD Countries, Curtin University, School of Economics and Finance [online]
http://espace.library.curtin.edu.au/R?func=dbin-jump-full&local_base=gen01-
era02&object_id=200881 (accessed 3 May 2015).
Stock, J.H. and Watson, M. (1993) ‘A simple estimator of cointegrating vectors in higher order
integrated systems’, Econometrica, Vol. 61, No. 4, pp.783–820.
Sulaiman, D.M. and Muhammad, U. (2010) ‘The bound testing approach for co-integration and
causality between financial development and economic growth in case of Pakistan’, European
Journal of Social Sciences, Vol. 13, No. 4, pp.525–531.
World Bank (2015) Countries and Economies [online] http://data.worldbank.org/country (accessed
10 March 2015).
York, R. (2007) ‘Demographic trends and energy consumption in European Union Nations:
1960–2025’, Social Science Research, Vol. 36, No. 3, pp.855–872.
York, R. and Rosa, E.A. (2012) ‘Choking on modernity: a human ecology of air pollution’, Social
Problems, Vol. 59, No. 2, pp.282–300.
York, R., Rosa, E.A. and Dietz, T. (2003a) ‘Footprints on the earth: the environmental
consequences of modernity’, American Sociological Review, Vol. 68, No. 2, pp.279–300.
York, R., Rosa, E.A. and Dietz, T. (2003b) ‘A rift in modernity? Assessing the anthropogenic
sources of global climate change with the STIRPAT model’, International Journal of
Sociology and Social Policy, Vol. 23, No. 10, pp.31–51.
Notes
1 According to econometric theory, obtained regression results are spurious if a linear
combination of non-stationary variables is not stationary, i.e., there is no cointegrating
relationship among them (see Engle and Granger, 1987; inter alia).
2 Half the distance to equilibrium will take 2.89 years [ln(0.5) / ln(1 + adjustment coefficient) =
ln(0.5) / ln(1 + (–0.213)) = 2.89].
3 Time series studies for Kazakhstan have been preferred to review. In the case of absence of
such studies, cross-sectional and or panel studies for Kazakhstan have been reviewed.
4 Dependent variables of these studies were either CO2 emission or ecological footprint.
5 The reason for running two different specifications (i.e., not including the both age groups in
one specification) is small number of observations.
