Impact of Climate Change On Sectoral Electricity Demand in Turkey

Energy Sources, Part B: Economics, Planning, and Policy
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/uesb20
Impact of climate change on sectoral electricity

demand in Turkey
Denizhan Guven, M. Ozgur. Kayalica, Gulgun. Kayakutlu & Erkan. Isikli
To cite this article: Denizhan Guven, M. Ozgur. Kayalica, Gulgun. Kayakutlu & Erkan. Isikli
(2021): Impact of climate change on sectoral electricity demand in Turkey, Energy Sources, Part B:
Economics, Planning, and Policy, DOI: 10.1080/15567249.2021.1883772
To link to this article: https://doi.org/10.1080/15567249.2021.1883772
Published online: 10 Mar 2021.
Submit your article to this journal
Article views: 78
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=uesb20
ENERGY SOURCES, PART B: ECONOMICS, PLANNING, AND POLICY
https://doi.org/10.1080/15567249.2021.1883772
Impact of climate change on sectoral electricity demand in Turkey

a b c d
Denizhan Guven , M. Ozgur. Kayalica , Gulgun. Kayakutlu , and Erkan. Isikli
a
Eurasia Institute of Earth Sciences and TEGAM, Istanbul Technical University, Istanbul, Turkey; bManagement
Engineering Department and TEGAM, Istanbul Technical University, Istanbul, Turkey; cEnergy Institute and TEGAM,
Istanbul Technical University, Istanbul, Turkey; dIndustrial Engineering Department, Istanbul Technical University,
Istanbul, Turkey
ABSTRACT KEYWORDS
This paper analyzes the impact of climate change on Turkey’s residential, Underlying Energy Demand
industrial and commercial electricity demand by applying the Structural Trend (UEDT); electricity
Time Series Model. The model consists of explanatory variables and uses Demand; structural Time
annual time series data for the period from 1980 to 2017. The variables Series Model (STSM); climate
Change; space Cooling in
included are electricity price, income per capita, urbanization rate, manufac Buildings
turing added value, population-weighted cooling degree days (as a climate
variable) and, finally, stochastic Underlying Energy Demand Trend. The effect
of cooling degree days on the residential, industrial and commercial electri
city demand is found to be significantly elastic. Moreover, residential, indus
trial and commercial electricity demands in 2031 are expected to increase
considerably, and at different ratios, depending on the chosen climate
change scenario. The findings of this study are significant for efforts to
establish regulations to shape energy security policies in the presence of
climate change.
1. Introduction
With a population of almost 82 million in 2018 (TurkStat 2018a), Turkey is located at the crossroads of
Asia and Europe. While only half of the country had access to electricity at the beginning of the 1970s,
almost everyone did by 1987 (Altas, Fikret, and Celebi 1994). In parallel with these developments, the
Turkish economy has become the 20th largest in the world (IMF 2018). Due to a growing population
and improving living standards, residential electricity need has been increasing by approximately 8.5%
annually, from 658 GWh to 54.3 TWh between the 1960s and 2017. The share of residential electricity
consumption has also risen from 16% to 22% in this period. Similarly, industrial electricity demand
increased from 4.7 TWh to 116.5 TWh between 1970 and 2017 (TEDC 2018).
Although the literature on forecasted energy demand for Turkey is vast, it is somewhat limited, as
these studies ignore the impact of climate change. There is, admittedly, literature on forecasted
electricity demand for other countries that includes the impact of climate change by using Cooling
Degree Days (CDD) as an explanatory variable. Yet CDD data does not properly represent electricity
consumption. Additionally, there is a growing body of literature on electricity demand forecasting
using exogenous factors such as energy efficiency developments, changes in behaviors, consumer
tastes, legislation and more, referred to as the Underlying Energy Demand Trend (UEDT). However,
the research on Turkey is both limited and narrow-angled, as it considers only one sector of the
economy at a time. In fact, as will be discussed in the first part of the next section, only residential and
industrial sectors are investigated – and these studies are conducted separately, in two different works.
This paper will therefore contribute to the literature by filling two gaps. First, it will attempt to
incorporate the impact of climate change simply by introducing the population-weighted cooling
CONTACT Denizhan Guven [email protected] Eurasia Institute of Earth Sciences, Avrasya Yer Bilimleri Enstitüsü, Istanbul
Technical University, Istanbul Teknik Universitesi, Ayazağa Kampüsü, Maden Fakültesi A Girişi, Sarıyer, İstanbul 34469, Turkey.
© 2021 Taylor & Francis Group, LLC
2 D. GUVEN ET AL.
degree days parameter into the forecast model. Thus, the paper will calibrate the CDD data simply by
focusing on city centers with high population densities, as opposed to uninhabited zones such as
forests. Second, the aforementioned UEDT will be implemented for i) residential, ii) industrial, and iii)
commercial sectors at the same time to provide a wider picture.
The quantitative analysis of electricity needs plays an important role in energy policies, including
the determination of energy policies and savings plans, capacity increases, long-term import contracts,
and planning of countrywide interconnected network construction. Turkey’s energy demand will
likely continue to increase in the future, and the country is in the process of constructing a number of
new power plants to meet the demand. Despite growing investments in renewable energy, the share of
CO2-based power plant investments remains high, as shown in Figure 1. As a result, CO2 and other
greenhouse gas (GHG) emissions increase year by year. According to the Intergovernmental Panel on
Climate Change (IPCC 2013), the global mean surface temperature observed between 2006–2015 was
approximately 0.87°C higher than in pre-industrial times (1850–1900). It is expected, except by the
most optimistic forecast, that global temperatures will increase more than 1.5°C compared to the pre-
industrial period before 2100. Even worse, many of the impacts of climate change will remain for
hundreds of years even if emissions disappear. Since 1990, CO2 emissions have skyrocketed globally,
and the trend only worsened into the 2000s and 2010s. Carbon dioxide concentrations have reached
415 ppm in May 2019, globally speaking, for the first time in human history (Scripps Institution of
Oceanography 2019).
Moreover, in parallel with climate change, global energy use for indoor cooling has doubled from
3.6 EJ to 7 EJ, and cooling is the fastest-growing energy use in buildings (Figure 2). In light of these
facts, the share of energy demand for cooling in buildings is expected to reach 14% of total electricity
demand in 2050 (IEA 2018a). This issue will be factored into our study using the “cooling degree days”
(CDD) metric, which measures the energy demand to cool buildings. Climate change significantly
increases CDDs, and a 1°C increase in global mean temperature would result in an approximately 25%
increase in average CDDs by 2050 compared with 2018 (IEA (International Energy Agency) 2018a).
This relationship underlines a major link between climate change and electricity consumption for
space cooling.
Efforts to mitigate climate change by reducing GHG, especially by reducing CO2 emissions, are
more important than ever. Under the COP21 agreement, Turkey has committed to reduce up to 21%
of its 2016 GHG (246 Mt CO2-eq) emissions by 2030 (UNFCCC 2015). But the Turkish government
announced in May 2019 that to improve energy security and meet its increasing energy demand, it will
80
70
Percentage (%)
60
50
40
30
20
10
0
2014 2015 2016 2017 2018 2019
Year
Wind Biogas/Biomass Thermal Hydro Solar

Figure 1. Share of energy investments in Turkey in terms of energy source (Source: prepared by the author, based on General
Directorate of Energy Affairs (2019) data).
ENERGY SOURCES, PART B: ECONOMICS, PLANNING, AND POLICY 3
2500
Share of total final energy

6.0
2000
TWh equivelents
4.5
1500
1000 3.0
500 1.5
0 0.0
1990 1995 2000 2005 2010 2015
Natural gas
Electricity
Share of energy use in buildings
Figure 2. Energy use for cooling in buildings (Source: IEA, 2018a).
encourage domestic coal use in power generation. According to TurkStat data, from 2010 to 2018, the
share of total production from thermal power plants has increased from 26.1% to 37.2%. During this
same period, meanwhile, the share of hydroelectric has decreased from 24.5 to 19.7 (TurkStat, 2019a).
Thus, efficient use of energy is critical for a country that meets almost 70% of its energy needs by
burning fossil fuels (IEA, 2018b). Furthermore, Turkey’s energy self-sufficiency is decreasing year
by year (Ozcan 2018). This creates a fragile economy due to dependence on foreign sources. The
motivation for this paper emerges exactly at this point. Determining key drivers of electricity demand
is crucial to preparing for regulations and policies that may shape future energy security, taking into
account ongoing climate change as well as exogenous factors.
This study is structured as follows. The literature background on electricity demand studies for
Turkey, as well as the effects of climate on electricity demand modeling, are reviewed in the next
section. In Section 3, the residential, industrial and commercial electricity demand forecasting models
that include economic and demographic indicators and climate change variables are developed. In
Section 4, the proposed models are run with appropriate data. In Section 5, the demand forecasting
models are run for three different scenarios, and all estimation results are discussed and compared
with previous studies. Finally, the last section concludes with an emphasis on possible policies and
regulations.
2. Background
This section provides a literature review on Turkish electricity demand, building on the background
information presented above on the official government forecast. This section also discusses literature
on the modeling approach used in this study. Previous studies that have employed the suggested
method are also covered briefly, along with previous research on the modeling of the climate’s effects
on electricity demand.
Turkey has limited energy resources to meet its rapidly increasing demand for energy, making it
necessary to use these resources efficiently. In the past, studies on Turkish energy demand forecasting
were mostly carried out by the State Institute of Statistics (SIS – now known as TurkStat), the State
Planning Organization (SPO) and the Ministry of Energy and Natural Resources (MENR) until 2000
(Ediger and Tatlidil 2002). Energy demand studies undertaken in Turkey can be divided into two
periods: before and after 1984. Analyses in the first period were undertaken in 1966, 1967, 1972 and
1979 by SPO, and in 1973, 1975, 1977 and 1978 by MENR. They were based on various best-fit curves
(Aybar, Saral, and Fikret 1986). However, the results of these forecasts far exceeded the actual energy
demand values.
4 D. GUVEN ET AL.
According to Ediger and Tatlidil (2002), 1989 was a milestone for Turkey in terms of incorporating
modern econometric techniques into future energy planning and forecasting studies. As recom
mended by the World Bank, MENR began to use the Model for Analysis of Energy Demand
(MAED) in medium- and long-term energy demand forecasting. In compliance with the results
obtained using MAED, MENR began optimum production-investment planning with Wien
Automatic System Planning (WASP III) (Ugur and Gençyılmaz 1990). MAED was performed by
MENR six times: in 1986, 1990, 1994, 1997 and 2005. However, these studies were criticized as being
biased to return higher values, as they used target values supplied by the government (see, for example,
Dilaver (2009), Akay and Atak (2007), Ediger and Akar (2007), Hamzacebi (2007), Ediger and Tatlidil
(2002)). Another problem with these studies was that the forecasts were generated using inadequate
and inconclusive data from past years (Ediger and Tatlidil 2002).
In addition to MAED, EFOM-12 C Mark I (Energy Flow Optimization Model) and Kouris’s
Correlation Model were utilized. EFOM-12 C Mark I was developed in 1984 by the Commission of
European Communities and was used for Turkey for the period from 1981 to 1985 (Ercab, Durmaz,
and Sivrioglu 1990). Kouris’s Correlation model was used to forecast Turkey’s primary and secondary
energy demands between 1970 and 1992 (Akdeniz and Demir 1994).
2.1. Previous electricity demand modeling for Turkey

This section reviews the literature on Turkey’s electricity demand modeling in two parts. The first will
provide an overview of the models focusing on residential electricity; the second will provide an
overview of those focusing on industrial electricity. Note, however, that some of the studies focus on
both.
Although the body of literature is not vast, several studies do examine residential electricity
demand modeling in Turkey. Nearly all use different models for different time periods. Hamzacebi
(2007) shows that the Artificial Neural Network (ANN) method is more accurate than the afore
mentioned MAED. Bilgili et al. (2012) also use ANN, but they compare it with the results of linear
regression (LR) and non-linear regression (NLR) instead, based on the installed capacity, gross
electricity generation, population and total subscribership indicators. According to this comparison,
the ANN model performs better than both the LR and NLR models, and also better than the
abovementioned prediction results of MENR. Halicioglu (2007) uses Bounds Testing to calculate
short- and long-term price and income elasticities depending on the number of chosen lags of
residential electricity demand in Turkey. Dilaver (2009) and Dilaver and Hunt (2011a) estimate
Turkish residential electricity demand for three different scenarios using the Structural Time Series
Model (STSM).
Arisoy and Ozturk (2014) estimated income and price elasticities for the period 1960–2008 with the
time-varying parameters method. Recently, Kilic and Ozdemir (2018) used the Long-range Energy
Alternatives Planning (LEAP) model, a scenario-based software tool, to analyze the economic, physical
and environmental effects.
There are also several studies in the related literature that model Turkey’s industrial electricity
demand. Ozturk et al. (2005) use Genetic Algorithms (GAs) to estimate industrial and total electricity
demand with population, Gross National Product (GNP and import and export data as indicators. The
study establishes two non-linear models, one in quadratic form and the other in exponential form, to
forecast electricity demand for 2002–2025. The forecast results of this study are higher than the
projection supplied by MENR. Using the same data as Hamzacebi (2007), Akay and Atak (2007) aim
to obtain both industrial and total electricity demand with the Grey Prediction with Rolling
Mechanism (GPRM) method. The authors predict an industrial electricity demand for 2015 of around
140.37 GWh. The forecast results are further analyzed for the projections of MENR and the authors
state that GPRM leads to more precise results than MENR. Dilaver and Hunt (2011b) estimate
industrial electricity demand with Structural Time Series Model (STSM) using industrial value
added and real industrial electricity price data.
Table 1. Residential and industrial electricity demand forecasting studies for Turkey.
Forecasted
Reference Method Data Variables years
Ozturk et al. Genetic algorithm electricity 1980–2001 GNP, population, import and export 2002–2020
(2005) demand model
Sozen, Akcayol, Artificial neural network (ANN) 1953–2000 Population, gross generation, installed capacity Model is
and established
Arcaklioglu
(2006)
Hamzacebi Artificial neural network (ANN) 1970–2004 Sectoral electricity consumption 2005–2020
(2007)
Halicioglu Bounds Testing approach 1968–2005 Residential electricity consumption per capita, real Model is
(2007) income per capita, real residential energy price established
index and urbanization ratio
Akay and Atak Grey prediction with rolling 1970–2004 Industrial electricity consumption 2006–2015
(2007) mechanism (GPRM)
Cunkas and Artificial neural networks (ANN) 1981–2007 GNP, GDP, population, number of households, 2008–2014
Altun (2008) industrial index, crude oil and electricity price
Dilaver and Structural time series model 1960–2008 Household total final expenditure, UEDT and real 2009–2020
Hunt (2011a) (STSM) residential electricity price
Dilaver and Structural time series model 1960–2008 Industrial value added, real industrial electricity 2009–2020
Hunt (2011b) (STSM) price and UEDT
Cunkas and Genetic programming 1988–2007 Electricity consumption 2008–2020
Taskiran (2011)
Bilgili et al. Comparison of Linear, Nonlinear 1990–2003 Installed capacity, gross electricity generation, 2008–2015
(2012) regression analysis and Artificial population and total subscribership
neural networks
Arisoy and A time varying parameter 1960–2008 Real income and real price Model is
Ozturk (2014) approach established
Gulcu and Particle Swarm Optimization 1979–2013 GDP, population, import and export 2014–2030
Kodaz (2017) algorithm
Kilic and LEAP Model 1970–2012 GDP-PPP, GDP-MER, population 2011–2023
Ozdemir (2018)
Uzlu (2019) Jaya algorithm – trained ANN 1980–2014 GDP, population, import and export 2015–2023
In Table 1 below, we provide a summary of the literature on residential, industrial and commercial
electricity demand forecasting, in which different methods, such as artificial neural networks, time
series models, particle swarm intelligence approach, bound testing approach and the LEAP model, are
all used. However, these studies all fail to consider the effects of climate change, one of the most
important phenomena of our time. This study therefore contributes to the existing literature by
predicting Turkey’s future residential, industrial and commercial electricity demand while taking
into account the impact of climate change.
2.2. Literature review of the modeling approach

Various methodologies and models with different specifications can be found in the energy economics
literature. One is the Structural Time Series Model (STSM) introduced by Harvey et al. (1986). Prior to
STSM, unit-root tests associated with cointegration techniques had been widespread in energy
demand forecast models for years (Hendry and Juselius 2000). Despite their frequent use, the
cointegration method and unit-root test have been criticized for their weak statistical characteristics
and unnecessary and/or misleading procedure (Harvey 1997; Maddala and Kim 1998). Seeking to
replace the cointegration method, Harvey (1997) combined the flexibility of time series models and the
interpretation of the conventional regression model. STSM is well-suited to unobservable trends
varying stochastically over time; thus, the combination of STSM and Autoregressive Distribution
Lag (ARDL) can produce energy forecast functions. This structure allows for both stochastic trends
and stochastic seasonality to determine price/income elasticities. Furthermore, the generalized model
discussed in Section 3 contains the key factors of income, price, urbanization and climate change, as
6 D. GUVEN ET AL.
well as a stochastic UEDT. The UEDT can be described as a factor for exogenous effects such as energy
efficiency developments, changes in behaviors, tastes and legislation. As a result, STSM/UEDT models
have been used to forecast energy demand by various studies, as shown in Table 2.
2.3. Effects of climate on electricity demand modeling

In acknowledgment of the growing effects of global warming, several scholars have examined the
impact of climate conditions on the consumption of energy. For the United States, Amato et al. (2005)
use a two-step estimation method to analyze the significance of climate change for both commercial
and residential energy demands. The results of this study indicate that residential energy demand is
highly sensitive to temperature. Thus, climate change may lead to a decrease in fuel and electricity
demand for heating in the winter and an increase in electricity demand in the summer.
Eskeland and Mideksa (2009) analyze the impact of weather conditions on electricity demand for
30 European countries. They demonstrate that the energy demand for heating in Northern Europe will
decline, while the energy demand for cooling in Southern Europe will rise alongside changes in the
climate. Furthermore, De Cian, Lanzi, and Roson (2013) present an empirical method to characterize
the relationship between temperature and residential energy demand. They also note that climate
change results in a clear heating effect for colder zones and both cooling and heating effects in hotter
zones.
The effect of climate change on the heating and cooling energy demands of residential buildings has
been investigated for Brazil (Invidiata and Ghisi 2016). According to their simulations, annual energy
demand is expected to increase by 19% to 65% through 2020, 56% to 112% through 2050, and 112% to
185% through 2080. However, they also estimate that the annual energy demand for heating will
decrease by about 94% by 2080 due to global warming. Similarly, Perez-Andreu et al. (2018) estimate
the impact of climate change on energy demand for cooling and heating homes in the Mediterranean
area. Their results show that heating energy demand for indoor spaces decreases when ambient
temperature rises. Moreover, in all scenarios, both the energy demand for cooling and the chance of
overheating increase with the climatic warming trend.
Ang, Wang, and Ma (2017) investigate the impact of increasing temperature on electricity con
sumption in Hong Kong and Singapore using two different approaches, namely the time series
approach and the decomposition approach. The authors show that an increase in average temperature
results in greater electricity consumption for both Singapore and Hong Kong. They also estimate that
the largest increase in consumption by percent would be in residential buildings, as opposed to
commercial and industrial sectors. Similar to previous research, Craig and Feng (2017) examine the
relationships between residential electricity consumption, climate variability, and the effects of energy
Table 2. Some selected STSM studies.

Reference Range of Countries Sectors
Harvey and Koopman (1993) Northwest US
Hunt, Judge, and Ninomiya (2003a) UK Aggregate and transport oil
Hunt and Ninomiya (2003) UK and Japan Transport
Hunt and Ninomiya (2005) Japan Aggregate
Amarawickrama and Hunt (2008) Sri Lanka Aggregate
Broadstock and Hunt (2010) UK Transport
Adeyemi et al. (2010) OECD Countries Aggregate
Sa’ad (2011) South Korea and Indonesia Aggregate and Residential
Dilaver and Hunt (2011a) Turkey Residential
Dilaver and Hunt (2011b) Turkey Industrial
Dilaver and Hunt (2011c) Turkey Aggregate
Adeyemi and Hunt (2014) OECD Countries Industrial
Javid and Qayyum (2014) Pakistan Aggregate
Atalla and Hunt (2016) GCC Countries Residential
Rodrigues, Losekann, and Filho (2018) Brazil Transport
efficiency programs for the United States. They show that there is a strong positive correlation between
fluctuations in climate and residential electricity consumption.
Another study shows that net expenditures for electricity are not homogeneous throughout the
globe (Clarke et al. 2018). In some cold territories such as Russia and Canada, net energy expenditures
would decrease under climate change, while in regions with more space-cooling demand, net energy
expenditures would increase.
Fan, Hu, and Zhang (2019) estimates the effects of climate on electricity demand in China using
a fixed-effect regression feedback model. The model also forecasts the future electricity demand for
three different climate change scenarios. This study shows that a 1% increase in cooling degree days
(CDD) would cause a 0.094% increase in electricity demand per capita. In one of the most recent
studies, D’Amico et al. (2019) aims to prove that estimating building energy demand depending on the
degree day is appropriate, on the condition that the assignation of the climate index is a part of the
same weather data.
Though it may seem that there has been an abundance of research on the topic, the impact of
climate change on electricity demand in Turkey has not been sufficiently analyzed. Using multiple
linear regression and ANN models, Gunay (2016) models the gross electricity consumption of Turkey
as a function of GDP per capita, population, inflation rate, unemployment percentage, and average
winter and summer temperatures. However, the climatic variables used are fairly rudimentary, as they
are calculated by taking the average of all meteorological stations throughout the country. Hence,
Gunay’s findings do not reflect the specific population distribution across the country. We overcome
this problem by developing a population-weighted cooling degree days model in Section 4. This helps
us produce an accurate assessment of the impact of climate change on Turkish electricity demand.
3. Methodology
In this section, we present the model setups. It is assumed that the residential electricity demand of
Turkey is identified by the following equation:
Er t ¼ f ðYt ; Pr t ; URt ; CDDt ; UEDT r t Þ (1)
where Ert Yt Prt
URt CDDt and UEDTrt represent residential electricity demand, real GDP per capita,
real residential electricity price, urbanization rate, population-weighted cooling degree days and the
underlying energy demand trend for residential electricity demand, respectively. Consumer theory
defines demand as a function of several parameters, the price of the good and income of the consumers
possibly being the most important ones. Hence, we include them as explanatory variables, as do other
studies in the literature. According to Holtedahl and Joutz (2004), electricity consumption expands
with urbanization, due to greater access to electricity and the need to purchase new electrical
equipment. Since Turkey has shown a quicker urbanization rate in recent years, especially during
the period we are focused on, we choose to include urbanization in the residential model. We also
include the climate change variable, CDD, which is often used to estimate climatization needs for
buildings. Finally, we include UEDT, which is explained in detail following Equation (7).
To estimate Equation (1) the dynamic autoregressive distributed lag specification is used as follows:
" #
X2
r r r
et ¼ γ1 et 1 þ γ2 et 2 þ αi yt i þ φi pt i þ δi URt i þ θCDDt þ UEDTtr þ εt
r
(2)
i¼0
where ert, yt, prt and cddt are natural logarithms of Ert, Yt, Prt and CDDt in each year, respectively.
While the coefficients α0, φ0, δ0 and θ0 stand for the short-term impact elasticities of income, price,
urbanization rate and cooling-degree-days respectively, εt represents a random white noise error term.
Furthermore, theQ long-term income, price and urbanization rate elasticities can be calculated by
φ þφ þφ
Γ ¼ α10 þαγ 1 þα2
, ¼ 10 γ 1 γ 2 , and Δ ¼ δ10 þδγ 1 þδγ 2 , respectively.
1 γ2 1 2 1 2
The industrial electricity demand of Turkey is assumed by the following equation:
8 D. GUVEN ET AL.
�
Ein t ¼ f Mt ; Pin t ; CDDt ; UEDT in t (3)
in in in
where E t, Mt, P t, CDDt, UEDT t represent industrial electricity demand, manufacturing value-
added, real industrial electricity price, population-weighted cooling degree days and Underlying
Energy Demand Trend for industrial electricity demand, respectively. The rationale for choosing
these variables is provided after Equation (1). The same explanation applies here with one difference:
manufacturing value-added is used instead of GDP, simply because while GDP can be used as a proxy
for understanding the economy’s overall performance, it does not say much about the industry.
Manufacturing value-added, on the other hand, provides a better measure for industrial development.
Urbanization rate is calculated as the ratio of household members living in city centers to the total
population. Since it is based on place of residence, this indicator is only used for residential electricity
demand, and is thus excluded here.
The dynamic autoregressive distributed lag specification equation is formed to evaluate Equation
(3) as follows:
" #
X
2
ein in in
t ¼ π 1 et 1 þ π 2 et 2 þ ζ i mt i þ υi pin in
t i þ λCDDt þ UEDTt þ εt (4)
i¼0
Unlike Equation (2), mt is the natural logarithm of manufacturing value-added. The long-term value-
added and price elasticities can be calculated by
Λ ¼ ζ10 þζπ11 þζπ22 and Φ ¼ υ10 þυ 1 þυ2
π1 π2 , respectively.
Commercial electricity demand in Turkey is calculated as follows:
�
Ecom t ¼ f Yt ; Pin t ; CDDt ; UEDT com t (5)
where Ecomt and UEDTcomt stand for commercial electricity demand and the Underlying Energy
Demand Trend for commercial electricity demand, respectively. The dynamic ARDL equation is given
in Equation (6):
" #
X
2
ecom
t ¼ τ 1 ecom com
t 1 þ τ 2 et 2 þ αi yt i þ υi pin com
t i þ κCDDt þ UEDTt þ εt (6)
i¼0
The long-term income and price elasticities for the commercial sector are
Ψ ¼ α10 þατ11 þα υ0 þυ1 þυ2
τ2 and Ω ¼ 1 τ 1 τ 2 , respectively.
2
Following Harvey et al. (1986), UEDT is a stochastic trend estimated by the STSM as:
�
μt ¼ μt 1 þ βt 1 þ ηt ; ηt ,NID 0; σ 2 η (7)
�
βt ¼ βt 1 þ �t ; �t ,NID 0; σ 2 � (8)
where notations μt and βt stand for the level and slope of the UEDT, respectively. The parameters ηt
and ξt are the mutually uncorrelated white noise disturbances with zero means and variances σ2η and
σ2ξ (known as hyperparameters). The terms ηt and ξt denote the shape of the stochastic trend
component (Harvey and Shephard 1993). Furthermore, to ensure the model passes the diagnostic
analysis (tests) for both standard and auxiliary (slope and level, irregular) residuals, it can sometimes
be essential to include outlier or irregular interventions (Irr), slope interventions (Slp) and/or level
interventions (Lvl) in the model. According to Harvey and Koopman (1992), these might provide
information on remarkable breaks and structural changes. Hence, the UEDT can be formulated as:
UEDTt = μt + irregular interventions + level interventions + slope interventions
For this reason, Equations (2) and (4) – (8) are determined by the combination of Kalman filter and
maximum likelihood. In addition to eliminating insignificant variables, interventions are added to
ensure the model passes an array of diagnostic tests. To estimate the proposed model, the software
package program STAMP 8.3 (Koopman et al. 2007) is used.
4. Data and estimation

The data used in this research and the estimation results depending on this collected data will be
presented in this section.
4.1. Data
Data used in this research was collected from various sources. The annual residential, industrial and
commercial electricity consumption (MWh) data from 1980 to 2017 was obtained from the Turkish
Electricity Distribution Corporation (TEDC 2018), which publishes electricity consumption data
(Wh) every year. In this report, electricity consumption is given as “Total electricity consumption,”
and the shares of each sector are represented as percentages. This data was generated by multiplying
total electricity consumption (MWh) by the share of residential electricity consumption in total
consumption; a similar calculation is performed for industrial and commercial electricity consump
tions. Time series for the real GDP per capita ($), the urbanization rate of Turkey and the real
manufacturing value added ($) were obtained from the World Bank, with 2010 as the base year (2018a,
2018b, 2018c). Real residential and industrial electricity prices ($/MWh) by year were obtained from
the International Energy Agency (IEA 2019a). In Turkey, prices for all types of electricity are constant
throughout the country and are determined by the Energy Market Regulatory Authority (EMRA).
Thus, electricity prices are taken as published by IEA.
Finally, we created a population-weighted CDD. The annual population-weighted CDD data of the
selected city centers for the 1980–2017 period was gathered from NASA maps with 25 km spatial
resolution (NASA 2015). However, to better represent the impact of climate on electricity demand, we
selected CDD data for city centers with high population densities instead of uninhabited zones such as
forests, lakes and cultivated areas. The population-weighted CDD data was generated using Equation (9):
P20
PPLi � CDDit
CDDt ¼ i¼1P20 t i
(9)
i¼1 PPLt
where PPLti and CDDti stand for the population and cooling degree day of city i for year t, respectively.
Most of Turkey’s populous and industrialized cities (altogether containing more than 65% of the
country’s total population) were selected in order to calculate CDDt more accurately. The populations
of selected cities were taken from TurkStat, and the missing data was interpolated linearly from the
available data (TurkStat 2019b). For the CDD calculation, the inclusion of only 65% of the population
might initially appear to be a flaw, but in truth the influence of the remaining cities on CDD is rather
limited. Thus, only the cities that affect CDD excessively were taken into consideration. Additionally,
since the use of natural gas and coal for space heating in buildings is promoted by the Turkish
government, coal, oil and natural gas are the main energy sources for space heating. Since the share of
electricity used for heating is almost zero, the impact of heating on electricity demand was omitted
from the models (IEA 2019b). Lastly, the brief statistical analysis of all variables is shown in Table 3.
4.2. Estimation results

During the estimation stage, four different specifications are tested, including stochastic levels/
stochastic slope, stochastic level/fixed slope, stochastic level/no slope and fixed level/fixed slope. The
preferred specification (stochastic level/stochastic slope) passes all diagnostic tests for residuals and
auxiliary residuals. Test results are presented in Table 4.
4.2.1. Residential electricity

While the slope and level components have statistically significant coefficients, their hyperparameters
are found to be zero – that is, the hyperparameters of the slope and the level components do not vary
over time. Thus, we are left with a conventional deterministic linear trend model that is no more than
10 D. GUVEN ET AL.
Table 3. Brief statistical analysis of variables.

Variable Min q1 Mean Median q3 Max St. Dev. IQR
Ein 13.01 27.83 51.46 46.32 72.99 116.54 29.76 45.16
Eres 4.39 8.63 23.4 21.31 38.53 54.29 15.93 29.9
Ecom 1.14 2.36 14.11 7.96 23.76 49.31 14.41 21.4
Y 1247 2084 5385 3420 9541 12543 3935.22 7457.38
Pin,com 48.65 71.62 91.88 84.82 106.00 150.92 28.94 34.38
Pres 35.27 62.76 98.01 86.24 120.82 189.96 46.42 58.06
CDD 699 747.5 812.8 818 878.5 923 72.87 131
UR 43.78 58.23 62.82 63.96 69.5 74.64 8.49 11.28
M 10832 25789 64661 40629 109695 156868 46972.28 83906.64
a restricted alternative of the general stochastic trend models. In other words, the UEDT is determi
nistic rather than stochastic. Real GDP per capita, real residential electricity price, urbanization rate,
and population-weighted cooling degree days are found to be the significant drivers of residential
electricity demand. The estimated model can be written as follows:
ert ¼ 0:3766ert 1 þ 0:1256yt þ 0:1019yt 1 0:1263prt 0:197URt þ 0:1898URt 1 þ 5:3971CDDt

þ UEDTtr þ εt
(10)
The test statistics and residual diagnostics (Table 4) for this model suggest a good fit, both to the data
and to the assumptions of the model. Note that additional normality tests for auxiliary residuals are
not generated, since both the level and the slope residuals are zero for the type of model estimated.
There are no autocorrelation problems, with all specifications passing the diagnostics tests presented.
The Durbin-Watson statistic is very close to 2 and r(1) is close to 0. Heteroscedasticity is not an
issue, as H ð9Þ ¼ 2:4249 is smaller than the statistical threshold F0:05;9;9 ¼ 3:18. Evidence of non-
normality is not found, since the corresponding statistics (1.3690) falls within the 95% confidence
intervalð0:051; 7:378Þ. R2d is non-negative, indicating that the model is distinctly better than random
walk with drift.
One might argue that conventional regression with a constant lack of trend could have also been
considered as an alternative to the one given above, as the parameter estimate for the slope component
is only significant at α ¼ 0:10. However, this would be impractical, since the time series used for
estimation is in logs, and thus a relatively small decrease in this logged series would correspond to
a larger decrease in the original series. The period between 2002 and 2017 is used to test the reliability
of the forecast model. The prediction test results and the CUSUM test are shown in Figure 3.
In conclusion, the white noise residuals and the prediction tests indicate that the model is correctly
specified. The specification with a deterministic trend and a constant is preferred by the residential
electricity demand data.
The estimated UEDT is mildly upward-sloping, as shown in Figure 4; thus, the UEDT is positive.
Since the UEDT is not stochastic, the average annual growth rate is constant. Energy price decreases
from 1973 to 1988, increases between 1988 and 2013, and decreases again in the period between 2013
and 2017. Therefore, in the first and the third periods, there is a chance that the price elasticity will be
under-estimated if the UEDT is ignored (Hunt, Judge, and Ninomiya 2003a). Similarly, failure to
incorporate the UEDT will result in an over-estimation of price elasticity in the second time period.
The estimated UEDT has a clear upward shape over the entire period of the study, indicating that the
UEDT in the residential electricity demand increased continuously even after controlling for income
and price effects, among others. On the other hand, as per Welsch (1989) and Hunt, Judge, and
Ninomiya (2003a), the lower price elasticity obtained might also indicate that a deterministic trend is
a reasonable alternative.
Table 4. Estimated results for residential, industrial and commercial electricity demand of Turkey.
Residential Industrial Commercial
Estimated Coefficients
γ1 0.37661* - -
α0 0.12557* - 0.21903*
α1 0.10193* - -
φ0 −0.12634* - -
δ0 −0.19726* - -
δ1 0.18980* - -
θ 5.39709* - -
ζ0 - 0.54786* -
υ0 - - −0.14721*
υ1 - −0.14538* -
λ - 3.25134* -
κ - - 4.55087**
LR Elasticities
Γ 0.365 - -
Π −0.203 - -
Δ −0.012 - -
Λ - 0.548 -
Φ - −0.145 -
Ψ - - 0.219
Ω - - −0.147
Hyperparameters
Irregular 0.0004 0.0 0.0
Level 0.0 0.0005 0.0016
Slope 0.0 0.0 0.0
Interventions
Lvl(1982)* 0.13998 Lvl(1987)* 0.10956 Lvl(1996)* 0.21095
Outlier(1983)* −0.05217
Goodness of fit
p.e.v. 0.0002904 0.000548 0.001434
AIC −7.5653 −7.1406 −6.1792
R2 0.99966 0.99873 0.99916
Rd2 0.88325 0.7731 0.67209
Residual Diagnostics
Std. Error (%) 1.7041 2.34 3.78
Normality 1.3690 0.26173 1.5040
Skewness −0.1215 −0.1215 0.36993
Kurtosis −1.0760 −1.0760 −1.2212
H(h) H(9) = 2.4249 H(10) = 0.65767 H(11) = 0.54763
r(1) −0.0777 0.031732 0.089426
r(p) r(6) = 0.2957 r(5) = −0.0415 r(6) = −0.0511
DW 2.1030 1.9205 1.7057
Q(p,d) Q(6,6–2) = 7.3635 Q(5,5–1) = 7.8337 Q(6,6–2) = 2.0693
Notes for Table 4:
(i) *, **Denotes statistical significance at 1% and 10%, respectively.
(ii) p.e.v. is the prediction error variance and AIC is the Akaike information criterion;
(iii) R2 is the coefficient of determination and R2d is the coefficient of determination based on differences, a measure that
benchmarks the estimated model against the random walk with drift. Mathematically, R2d ¼ PT 1 SSE 2 , where SSE is short
t¼2
ðΔyt ΔyÞ
for Sum of Squared Errors and Δy is the mean of the first differences (Harvey 1989).
(iv) DW is the Durbin-Watson statistic;
(v) r(1) is the estimated residual autocorrelations at lag 1 distributed approximately as N(0, 1/T)
Normality is the Bowman-Shenton statistic of the third and fourth moments of the residuals that is approximatelyχ 22 .
H(h) is a measure of heteroskedasticity that follows approximately Fh,h.
Q(p,d) is the Box-Ljung statistic depending on the first p autocorrelations that follow a Chi-square distribution.
4.2.2. Industrial electricity

Coefficients for both the slope and level components are found to be statistically significant in the case of
industrial electricity demand; however, only the hyperparameter for level is non-zero. The hyperpara
meter of the level component varies over time, indicating a local level model with drift; in other words, the
UEDT is now stochastic. Real manufacturing added value, real industrial electricity price and population-
12 D. GUVEN ET AL.
Figure 3. Prediction test results (a) and CUSUM test (b) for the stability.
Figure 4. Estimated Underlying Energy Demand Trend for residential electricity demand.
weighted cooling degree days are found to be the key drivers of industrial electricity demand. The
industrial electricity demand model can be written as follows based on the test results shown in Table 4:
ein
t ¼ 0:5479mt 0:1454pin in
t 1 þ 3:2513cddt þ UEDTt þ εt (11)
The test statistics and residual diagnostics (Table 4) for this model also suggest a good fit, both to the
data and to the assumptions of the model. There are no autocorrelation problems, since the Durbin-
Watson statistic is close to 2 and r(1) is very close to 0. Heteroscedasticity is not an issue either
(H ð10Þ ¼ 0:65767 < F0:05;10;10 ¼ 2:98). Evidence of non-normality is absent, since the corresponding
statistics (0.26173) falls within the 95% confidence interval ð0:051; 7:378Þ; however, note that this
value is quite close to the lower bound of the given interval. R2d is non-negative, but it is smaller than
that of the model for residential electricity demand.
To test the reliability and stability of the forecasting model, data for 2009–2017 is used, and the
prediction test results and CUSUM test are given in Figure 5.
In conclusion, the white noise residuals and the prediction tests indicate that the model is correctly
specified. The specification with a stochastic trend and a constant is preferred by the industrial
electricity demand data.
Figure 6 shows the estimated UEDT for industrial electricity demand. Its behavior is non-linear,
with a negative increasing trend throughout the estimation period. In this period, Turkey’s economic
state was fluctuating due to economic crises, tax reductions/increases, and so on; thus, this figure
accurately reflects the unstable situation.
Though it fluctuates, the estimated UEDT in this case is upward sloping (Figure 6); thus, it is
positive. Since it is also stochastic, the UEDT has periods of both upward and downward slope.
Therefore, the average annual growth rate is not constant. The behavior of the estimated UEDT over
the entire period of the study indicates that the UEDT for industrial electricity demand did not incline
continuously after controlling for price effects (and others). Nevertheless, increasing UEDT is an
indication of the increasing intensity of the industry over the last three decades.
4.2.3. Commercial electricity

As with industrial electricity demand, the statistical results also suggest a local level model with drift
when forecasting commercial electricity demand. Thus, the UEDT in this case is also stochastic. Real
industrial electricity price, real GDP per capita and population-weighted cooling degree days are
determined to be key components of commercial electricity demand, which is expressed in the
following equation:
ecom
t ¼ 0:21903yt 0:14721pcom
t þ 4:551cddt þ UEDTtcom þ εt (12)
The test statistics and residual diagnostics (Table 4) for this model suggest a fair fit, both to the data
and to the assumptions of the model. Autocorrelation does not appear to be an issue, since the Durbin-
Watson statistic (1.7057) is still close to 2 and r(1) is also close to 0. However, unlike with previous
results, some might suspect positive autocorrelation here. Nevertheless, the estimation results in this
case should be approached with care. Heteroscedasticity is not an issue since
H ð11Þ ¼ 0:54763 < F0:05;10;10 ¼ 2:82). Evidence of non-normality is absent, since the corresponding
Figure 5. (a) Prediction test results and (b) CUSUM stability test.
Figure 6. Estimated Underlying Energy Demand Trend for industrial electricity demand.
14 D. GUVEN ET AL.
Figure 7. (a) Prediction test results and (b) CUSUM stability test.
statistics (1.5040) falls within the 95% confidence interval ð0:051; 7:378Þ. R2d is still non-negative, but it
is prominently smaller now than that of the model for residential electricity demand.
To test the reliability and the stability of the model, prediction and CUSUM tests are performed.
Their results are shown in Figure 7.
Figure 8 shows the UEDT of commercial electricity demand. It has a negative increasing trend,
similar to that of industrial electricity demand. However, compared with industrial electricity demand,
the trend of commercial demand is smoother.
As with residential electricity demand, the estimated UEDT here is mildly upward sloping, as
shown in Figure 8; thus, the UEDT is positive. Since the UEDT is not stochastic, the average annual
growth rate is constant. The generally upward trend of the estimated UEDT over the period of the
study indicates that the UEDT in the commercial electricity demand inclined continuously even after
controlling for income, price and climate effects. Therefore, one can conclude that the sector has
become more electricity energy-intensive in Turkey over the last three decades.
4.3. Forecasting assumptions

Proper forecasting of future electricity demand is essential for Turkey, a country with limited energy
resources. Based on the NASA (2015) data, the estimated annual increase of population-weighted
CDD for Turkey is 0.5%. Thus, this rate was taken as the reference point. To create different forecast
outcomes, we assumed that low- and high-end scenarios would be half and twofold of the reference
scenario, respectively. Thus, considering the influence of climate change on electricity demand, the
increase in Turkey’s cooling degree-days is expected to be 0.5% (reference case), 0.25% (low case) or
1.0% (high case) annually. Additionally, residential/industrial electricity price, urbanization rate,
manufacturing value added and GDP per capita are extrapolated from the trend of the last 15 in-
sample observations due to the lack of future projections for these variables in Turkey. CDD is the only
variable that we use to create different scenarios. Hence, we distinguish CDD simply by assuming, in
our case, three different scenarios based on three different values.
5. Results and discussion

Given the above assumptions, the residential electricity demand of Turkey in 2031 will reach 151.2
TWh, 99.3 TWh and 80.1 TWh in the “high”, “reference” and “low” scenarios, respectively. Similarly,
industrial electricity demand will expand to 286.6 TWh, 222.4 TWh and 195.5 TWh in the “high”,
“reference” and “” scenarios, respectively. Commercial electricity demand will be 224.9 TWh, 193.5
TWh and 179.2 TWh in the “high”, “reference” and “low” scenarios, respectively. The results of annual
Turkish residential, industrial and commercial electricity demand forecast for the 2018–2031 period
for all three scenarios are shown below in Figure 9.
From the general equation of Turkish residential electricity demand (Equation 6), long-term
income elasticity and price elasticity are calculated to be 0.365 and −0.203, respectively. Long-term
Figure 8. Estimated Underlying Energy Demand Trend for commercial electricity demand.
(a) (b)
255000
205000
GWh
155000
105000
55000
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
Year
Low Ref High
(c)
Figure 9. (a) Residential (b) Industrial (c) Commercial electricity demand forecast for 2018–2031 Period.
16 D. GUVEN ET AL.
price elasticity is found to be little more than the values of −0.52 (estimated by Halicioglu (2007)),
−0.56 (estimated by Dilaver (2009)), and −0.38 (estimated by Dilaver and Hunt (2011a)). Since this
study seeks to forecast long-term electricity demand, it focuses on long-term price and income
elasticities. When compared with the above-mentioned literature on Turkish residential electricity
demand, two observations stand out. First, without exception, short-term elasticity values for price
and income are in parallel with those found in the literature. Second, price is found to be less elastic in
this study. In other words, electricity consumption is less price-sensitive; a price increase reduces
electricity consumption proportionately less than is found by other studies. Nevertheless, as is clear
from the figures above, our result is in parallel with the literature that demonstrates price inelasticity.
In contrast, the long-term income elasticity estimated by this study (0.365) is slightly smaller than the
values of 0.955 (reported by Arisoy and Ozturk (2014)) and 0.70 (reported by Halicioglu (2007)). Still,
the impact of income on electricity consumption shows a similar pattern that marks electricity as
a normal good.
Considering CDDt as the climate variable, the effect of climate on residential electricity demand is
significantly elastic, with a value of 5.39709. This elasticity falls between the values for the Middle East
(9.3) and Europe (1.2) found by IEA (2018a). However, no specific value had been found for the share
of energy consumption for space cooling (as a percentage of total building final-energy-use) in Turkey.
This study finds that share to be 5.4% for 2017. This means that in parallel with the IEA report (IEA
(International Energy Agency) 2018a), climate conditions affect residential electricity demand on
a vast scale: a 1% increase in the number of CDD increases residential electricity consumption by more
than 5%. Note that the IEA report focuses on regions and does not study Turkey separately; nor does it
investigate sectors, as we do here.
Unlike Halicioglu (2007), we find that the urbanization rate has a very small positive influence, with
the value of −0.012, on residential electricity demand in the long run. The negative relationship
between urbanization rate and residential electricity demand can be accounted for in three ways. First,
urbanized areas have more developed control mechanisms to prevent the illegal use of electricity. In
the eastern part of Turkey – where the urbanization rate is relatively low – there is massive illegal
electricity usage. For instance, comparing the eastern Anatolia region to the Marmara region, the
urbanization rates are 77% and 98%, respectively, while the illegal electricity usage rates are on average
38.4% and 5.7%, respectively (EMRA 2018; TurkStat 2018b). Control mechanisms impel citizens to
pay their electricity bills instead of using illegal electricity, thus reducing energy use. Second, urbani
zation enhances opportunities and increases education level. Better education, and the increased
awareness it brings, encourages people to become more conscious about energy efficiency. An energy-
conscious society reduces its electricity consumption by purchasing more efficient household equip
ment (such as A+++ refrigerators and washing machines). Finally, the share of buildings with proper
insulation is much higher in urbanized areas than in rural areas. Insulation contributes to better
energy efficiency in buildings, and adjoining buildings also use heating and cooling systems more
effectively.
The preferred model includes a level of intervention for 1982. This likely reflects that electricity
distribution was completely transferred to the Turkey Electricity Authority in 1982. This situation may
contribute to an increase in access to electricity in Turkey.
Considering the general equation of industrial electricity demand, long-term price elasticity is
calculated to be −0.145, and the long-term manufacturing added value to be 0.548. The long-term
price elasticity is quite close to the value of −0.15 estimated by Dilaver and Hunt (2011b). However,
our long-term price elasticity is slightly smaller than the value of −0.014 reported by Arisoy and
Ozturk (2014).
The impact of the climate variable “CDD” on industrial electricity demand is found to be highly
elastic, with a value of 3.25. This means that a 1% increase in CDD will cause approximately a 3.25%
increase in industrial electricity demand. Thus, as with residential electricity, climate conditions
greatly influence industrial electricity demand.
According to the general equation of commercial electricity demand, the long-term price and
income elasticities are calculated to be −0.147 and 0.219, respectively. Long-term price elasticity is thus
nearly equivalent to that of the industrial sector. Long-term income elasticity, meanwhile, is slightly
smaller than that of the residential sector. The impact of climate on commercial electricity demand is
estimated at 4.551 for 2017, again highlighting how strongly climate conditions affect commercial
electricity demand.
Since this paper analyzes sectoral electricity demands in Turkey, it mainly focuses on and compares
its results with studies specific to Turkey. Although the literature uses different methodologies and
scenarios for other countries/regions, it is worthwhile and interesting to compare the elasticities for
Turkey to those for other economies such as the United States, the European Union, China and
Indonesia. To accomplish this, a small group of samples from the literature on the relationship
between electricity demand and elasticities for different economies is gathered for comparison.
Apart from price elasticity for electricity demand in China, the price and income elasticities found
in all the studies mentioned below are less than 1; in other words, price and income are inelastic, and
electricity demand is less sensitive to changes in price and income.
Numerous studies explore sectoral electricity demand elasticities for the European Union. Two are
chosen here for comparison with this study. While the price elasticities for the residential and
industrial sectors in Turkey are similar to those found by Cialani and Mortazavi (2018), they are
smaller compared to those found by Csereklyei (2020). The income elasticities for both sectors in
Europe are nearly equivalent, while they are less sensitive in Turkey. This could suggest that relatively
higher income elasticity for European consumers compared to Turkish ones is a sign of greater
consumption of electrical appliances (e.g., tumble dryers), and thus electricity overall. This could be
due to climate conditions, cultural differences, differences in income levels or a combination of all
these factors. The difference in income elasticities is smaller between the EU and Turkey for the
industrial sector, but larger for the residential sector.
Price elasticity for residential electricity demand is considerably higher in China than in Turkey, the
US, the EU and Indonesia. One reason discussed in these studies is model specification and data, while
the other reason seems to be behavioral/economic factors, such as cutting back the use of household
appliances when electricity prices increase. In terms of income elasticity, the responsiveness of
demand for electricity to income change is negligible. However, the reason for this is not thoroughly
discussed by the authors. Furthermore, the paper is limited to residential electricity demand only.
In the US, the figures clearly show that the price elasticity of electricity demand in all sectors is
higher than in any other country, with the exception of the price elasticity for residential electricity
demand in China. The US is the only country in Table 5 with a positive price elasticity in the industrial
sector. This could imply that the industrial sector in the US has various alternatives to electricity such
as micro grids, co-generation, etc.
Increases in space cooling demands in buildings leads to a higher level of electricity production.
Producing electricity to meet this demand causes more CO2 emissions to be released into the air. As
mentioned earlier, this situation creates a greenhouse effect on a global scale. Because GHGs accelerate
climate change and global warming, space cooling demand will naturally increase further. This process
creates a vicious cycle, as shown in Figure 10.
Total electricity consumption for space cooling in Turkey in 2017 is calculated as 8.9 TWh (2.9
TWh from the residential sector, 3.8 TWh from the industrial sector, 2.2 TWh from the commercial
sector) for Turkey. By the end of 2017, more than 847.8 USD million was spent in Turkey to pay
electricity bills solely for space cooling. This electricity expenditure corresponds to about 0.1% of
Turkey’s GDP. With the increasing impact of climate change on electricity consumption, this rate is
expected to grow in the future. Furthermore, to produce this quantity of electricity, approximately 4.69
Mton CO2-eq emissions were released into the atmosphere in 2017. The social cost of CO2 emission is
taken as 4.46 USD/ton CO2 for Turkey (Ricke et al. 2018); thus, the social cost of emissions due to
electricity generation for space cooling is calculated to be more than 20.9 USD million.
18 D. GUVEN ET AL.
Table 5. Comparison of price and income elasticities.
Region/ Residential Industrial Commercial

Reference Country Price Income Price Income Price Income
This study Turkey −0.203 0.365 −0.145 0.548 −0.147 0.22
Cialani and Mortazavi (2018) EU −0.189/- 0.862/ −0.118/- O.740/
0.302 0.913 0.198 0.64
Csereklyei (2020) EU −0.53/-0.56 0.61 −0.75/-1.01 0.76/1.08
Burke and Abayasekara (2017) US −1.2 −0.3/-
0.6
Alberini, Gans, and Velez-Lopez US −0.86 0.02
(2011)
Sun (2015) US −0.65/-0.997 0.23/0.66
Shi, Zheng, and Song (2012) China −2.477 0.058
Al Irsyad, Nepal, and Halog (2018) Indonesia −0.2 0.56 0.24 0.22
Energy policy has long been a difficult issue for many developing countries, and Turkey is no
exception. Due to growing electricity demand, the country should establish regulations and policies
that can promote future energy security while considering the impact of climate change. Policy
recommendations can be channeled in three directions: i) residential, ii) industrial and commercial,
and iii) utility. For the residential sector, one of the biggest issues in Turkey is the illegal use of
electricity, as mentioned above. Non-technical losses hint at problems with social, economic, geo
graphic and cultural components. No country with illegal electricity use problems can expect dis
tribution companies to deal with electricity theft on their own. In the fight against illegal electricity use,
it is of great importance that the government supports the distribution companies with regulations
and enforces them with its institutions. Strengthening control mechanisms and laws – especially in the
eastern part of Turkey – may, ceteris paribus, lead to a decrease in total electricity consumption, as
mentioned in the previous section. Furthermore, distribution companies can prevent illegal electricity
use by installing smart electricity meters on the outside of residences or panels on power poles.
Another problem is electricity demand for space cooling in buildings. Bearing in mind that space
cooling is the fastest-growing energy use in buildings, the government should renew energy
performance codes for buildings. For new building construction, the government should carefully
inspect insulation and air-conditioning calculations and implementations. Moreover, the use of
Figure 10. Vicious circle of space cooling demand.

some state-of-the-art technologies, such as low emissivity coatings for ‘cool roofs,’ aerogel and
vacuum insulation panels (VIP) for insulation, should be widely adopted by construction and
architecture firms. The high cost of these materials should be subsidized by the government
and funded with relatively lower interest rates by financial institutions.
Household appliances account for a large portion of electricity consumption in residential build
ings, and inefficient appliances result in overuse of electricity. To prevent this problem, Minimum
Energy Performance Standards (MEPS) could be implemented by the government to limit the
maximum electricity consumption. As a result, the most commonly used appliances could be
regulated for minimum energy performance. To decrease the usage of less efficient equipment, the
government can put into practice a carbon taxation system. In this way, consumers would prefer to
switch to relatively more efficient equipment. Moreover, lighting is responsible for 15% of global
electricity consumption. Transitioning to energy-efficient lighting by decreasing the use of inefficient
incandescent lamps has great potential for reducing electricity demand in residential buildings.
Electricity distribution companies and the government could conduct a capacity-building effort to
inform household members about the benefits of replacing their incandescent lamps with high-
efficiency ones.
For the commercial and industrial sectors, there are many policies that could reduce electricity
demand. It has been argued that utilizing cleaner energy sources with more efficient technologies in
can decrease energy costs by up to 80% and increase energy savings by up to 30% (Salvarli and Salvarli
2017). To achieve this, the government could impose tax/subsidy incentives and make guarantee-of-
purchase agreements to boost renewable energy investments. Moreover, filtering systems must
become obligatory in all existing CO2-based power plants. Based on the results of this study, improv
ing the efficiencies of industrial and commercial equipment, providing capacity-building efforts to
inform consumers about energy efficiency measures and climate change mitigation strategies could all
help stabilize the growth rate of electricity demand.
Most importantly, the government should act to establish the Energy Efficiency Obligation Scheme
(EEOS) in Turkey as soon as possible. EEOS is a mechanism by which electricity distribution, supplier
and retailer companies providing electricity to end-users are responsible for implementing energy
efficiency action mechanisms for these users. Companies enrolled in this mechanism should be
rewarded or punished based on whether they meet a certain percentage of energy-saving targets or not.
6. Conclusion
This paper aims to estimate Turkish residential, commercial and industrial electricity demand given
GDP per capita, real electricity prices, urbanization rate, manufacturing added value, and cooling
degree-days. Using the STSM approach with data capturing the period from 1980 to 2017, the model
calculates the long-run income and price elasticities of sectoral electricity demands and obtains results
in parallel with the existing literature.
This study also compares the elasticities of Turkey to those of several selected countries/regions,
namely the US, EU, China and Indonesia. The results mostly imply that electricity demand in these
economies is proportionately less sensitive to changes in price and income.
The model results regarding residential UEDT indicate that technological development and
changes in consumer behavior, including population density, decreases in prices of household
electrical appliances and other factors, cause an increase in the residential sector’s UEDT. For
industrial and commercial electricity demand, the trend of UEDT increases, but not at a fixed rate.
The upward-sloping UEDT shows that the electricity efficiency of the manufacturing and commercial
sectors has decreased during the study period (or has increased but was outweighed by exogenous
factors such as politics).
Unlike previous electricity demand studies for Turkey, this study covers the impact of climate on
electricity demand by using population-weighted cooling degree day data. Above all, the main finding
of this study is that the impact of global warming and climate change on electricity consumption due
20 D. GUVEN ET AL.
to increased space cooling needs in buildings is very significant. Thus, by estimating future residential,
industrial and commercial electricity needs using structural time series tools, this study could assist the
government’s efforts in executing more efficient policies for energy security, while reducing CO2
emissions and meeting its COP21 and other climate commitments.
In future studies, countries could be clustered based on appropriate properties (e.g., latitude,
topography, economic indicators, etc.) so that their electricity demands for space cooling can be
properly studied. In this manner, the effect of climate on labor productivity, for instance, can be
compared and examined for such countries. Furthermore, the model used in this study could
be examined with different methodologies, e.g., a Bayesian network. The results of these models
could then be compared.
Acknowledgments
The authors would like to thank two anonymous reviewers and editor for their very helpful and strong suggestions.
ORCID
Denizhan Guven http://orcid.org/0000-0003-1605-7661
M. Ozgur. Kayalica http://orcid.org/0000-0001-9828-7385
Gulgun. Kayakutlu http://orcid.org/0000-0001-8548-6377
Erkan. Isikli http://orcid.org/0000-0002-8319-8782
References
Adeyemi, O. I., D. C. Broadstock, M. Chitnis, L. C. Hunt, and G. Judge. 2010. Asymmetric price responses and the
underlying energy demand trend: Are they substitutes or complements? Evidence from modeling OECD aggregate
energy demand. Energy Economics 32 (5):1157–64. doi:10.1016/j.eneco.2010.04.003.
Adeyemi, O. I., and L. C. Hunt. 2014. Accounting for asymmetric price responses and underlying energy demand trends
in OECD industrial energy demand. Energy Economics 45:435–44. doi:10.1016/j.eneco.2014.07.012.
Akay, D., and M. Atak. 2007. Grey prediction with rolling mechanism for electricity demand forecasting of Turkey.
Energy 32 (9):1670–75. doi:10.1016/j.energy.2006.11.014.
Akdeniz, H. A., and M. H. Demir. 1994. Enerji tüketiminin ekonometrik analizi ve Türkiye uygulaması [Econometric
analysis of energy consumption and application in Turkey]. Sixth Energy Congress of Turkey, Izmir, Turkey.
Al Irsyad, M., R. Nepal, and A. Halog. 2018. Exploring drivers of sectoral electricity demand in Indonesia. Energy
Sources, Part B: Economics, Planning, and Policy 13 (9–10):383–91. doi:10.1080/15567249.2018.1538271.
Alberini, A., W. Gans, and D. Velez-Lopez. 2011. Residential consumption of gas and electricity in the US: The role of
prices and income. Energy Economics 33 (5):870–81. doi:10.1016/j.eneco.2011.01.015.
Altas, M., H. Fikret, and E. Celebi. 1994. Türkiye enerji sektörünün gelişimi ve arz talep projeksiyonları (1970-2010)
[Development of Turkish energy sector and supply/demand projections (1970-2010)]. Sixth Energy Congress of
Turkey, Izmir, Turkey.
Amarawickrama, H. A., and L. C. Hunt. 2008. Electricity demand for Sri Lanka: A time series analysis. Energy 33
(5):724–39. doi:10.1016/j.energy.2007.12.008.
Amato, A. D., M. Ruth, P. Kirshen, and J. Horwitz. 2005. Regional energy demand responses to climate change:
Methodology and application to the commonwealth of Massachusetts. Climatic Change 71 (1–2):175–201.
doi:10.1007/s10584-005-5931-2.
Ang, B. W., H. Wang, and X. Ma. 2017. Climatic influence on electricity consumption: The case of Singapore and Hong
Kong. Energy 127:534–43. doi:10.1016/j.energy.2017.04.005.
Arisoy, I., and I. Ozturk. 2014. Estimating industrial and residential electricity demand in Turkey: A time varying
parameter approach. Energy 66:959–64. doi:10.1016/j.energy.2014.01.016.
Atalla, T. N., and L. C. Hunt. 2016. Modeling residential electricity demand in the GCC countries. Energy Economics
59:149–58. doi:10.1016/j.eneco.2016.07.027.
Aybar, E. N., R. Saral, and H. Fikret. 1986. Türkiye’nin bugünü ve gelecekteki enerji durumu [Energy today and in the
future in Turkey]. Fourth Energy Congress of Turkey, Izmir, Turkey.
Bilgili, M., B. Sahin, A. Yasar, and E. Simsek. 2012. Electric energy demands of Turkey in residential and industrial
sectors. Renewable and Sustainable Energy Reviews 16 (1):404–14. doi:10.1016/j.rser.2011.08.005.
Broadstock, D., and L. C. Hunt. 2010. Quantifying the impact of exogenous non-economic factors on UK transport oil
demand. Energy Policy 38 (3):1559–65. doi:10.1016/j.enpol.2009.11.039.
Burke, P. J., and Abayasekara, A. 2017. The Price Elasticity of Electricity Demand in the United States: A Three-
Dimensional Analysis. CAMA Working Paper No. 50/2017. Available at SSRN: https://ssrn.com/abstract=3016911
Cialani, C., and R. Mortazavi. 2018. Household and industrial electricity demand in Europe. Energy Policy 122:592–600.
doi:10.1016/j.enpol.2018.07.060.
Clarke, L., J. Eom, E. H. Marten, R. Horowitz, P. Kyle, R. Link, B. K. Mignone, A. Mundra, and Y. Zhou. 2018. Effects of
long-term climate change on global building energy expenditures. Energy Economics 72:667–77. doi:10.1016/j.
eneco.2018.01.003.
Craig, C. A., and S. Feng. 2017. Exploring utility organization electricity generation, residential electricity consumption,
and energy efficiency: A climatic approach. Applied Energy 185:779–90. doi:10.1016/j.apenergy.2016.10.101.
Csereklyei, Z. 2020. Price and income elasticities of residential and industrial electricity demand in the European Union.
Energy Policy 137:111079. doi:10.1016/j.enpol.2019.111079.
Cunkas, M., and A. A. Altun. 2008. Long term electricity demand forecasting in Turkey using artificial neural networks.
Energy Sources, Part B: Economics, Planning, and Policy 5 (3):279–89. doi:10.1080/15567240802533542.
Cunkas, M., and U. Taskiran. 2011. Turkey’s electricity consumption forecasting using genetic programming. Energy
Sources, Part B: Economics, Planning, and Policy 6 (4):406–16. doi:10.1080/15567240903047558.
D’Amico, A., G. Ciulla, D. Panno, and S. Ferrari. 2019. Building energy demand assessment through heating degree days:
The importance of a climatic dataset. Energy 242:1285–306.
De Cian, E., E. Lanzi, and R. Roson. 2013. Seasonal temperature variations and energy demand. Climatic Change 116
(3–4):805–25. doi:10.1007/s10584-012-0514-5.
Dilaver, Z. 2009. Residential electricity demand for Turkey: A structural time series analysis. In 10th IAEE European
Conference-Energy, Policies and Technologies for Sustainable Economies, Wien, Austria.
Dilaver, Z., and L. C. Hunt. 2011a. Modeling and forecasting Turkish residential electricity demand. Energy Policy 39
(6):3117–27. doi:10.1016/j.enpol.2011.02.059.
Dilaver, Z., and L. C. Hunt. 2011b. Industrial electricity demand for Turkey: A structural time series analysis. Energy
Economics 33 (3):426–36. doi:10.1016/j.eneco.2010.10.001.
Dilaver, Z., and L. C. Hunt. 2011c. Turkish aggregate electricity demand: An outlook to 2020. Energy 36 (11):6686–96.
doi:10.1016/j.energy.2011.07.043.
Ediger, V. S., and H. Tatlidil. 2002. Forecasting the primary energy demand in Turkey and analysis of cyclic patterns.
Energy Conversion and Management 43 (4):473–87. doi:10.1016/S0196-8904(01)00033-4.
Ediger, V. S., and S. Akar. 2007. ARIMA forecasting of primary energy demand by fuel in Turkey. Energy Policy 35
(3):1701–08. doi:10.1016/j.enpol.2006.05.009.
EMRA (Energy Market Regulatory Authority). 2018. Elektrik piyasası piyasa gelişim raporu 2017 [Electricity market
market development report 2017]. Accessed September 17, 2019. https://www.epdk.org.tr/Detay/
DownloadDocument?id=baMXxaZ/ZLs=[inTurkish].
Ercab, Y., A. Durmaz, and M. Sivrioglu. 1990. EFON-12 Enerji arz modelinin Türkiye’ye uygulanması [The application
of EFOM-12 energy demand model to Turkey]. Fifth Energy Congress of Turkey, Ankara, Turkey.
Eskeland, G. S., and T. K. Mideksa. 2009. Climate change and residential electricity demand in Europe. https://dx.doi.
org/10.2139/ssrn.1338835.
Fan, J., J. Hu, and X. Zhang. 2019. Impacts of climate change on electricity demand in China: An empirical estimation
based on panel data. Energy 170:880–88. doi:10.1016/j.energy.2018.12.044.
General Directorate of Energy Affairs. 2019. Electricity investments summary report. Accessed August 18, 2019. https://
www.eigm.gov.tr/tr-TR/Raporlar[inTurkish].
Gulcu, S., and H. Kodaz. 2017. The estimation of the energy demand using particle swarm optimization algorithm:
A case study of Turkey. 8th International Conference on Advances in Information Technology, IAIT2016, 19-22 Dec.
2016, China. Procedia Computer Science 111:64–70. doi:10.1016/j.procs.2017.06.011.
Gunay, M. E. 2016. Forecasting annual gross electricity demand by artificial neural networks using predicted values of
socio-economic indicators and climatic conditions: Case of Turkey. Energy Policy 90:92–101. doi:10.1016/j.
enpol.2015.12.019.
Halicioglu, F. 2007. Residential electricity demand dynamics in Turkey. Energy Economics 29 (2):199–210. doi:10.1016/j.
eneco.2006.11.007.
Hamzacebi, C. 2007. Forecasting of Turkey’s net electricity energy consumption on sectoral bases. Energy Policy 35
(3):2009–16. doi:10.1016/j.enpol.2006.03.014.
Harvey, A. C. 1989. Forecasting structural time series models and the Kalman filter. Cambridge: University Press.
Harvey, A. C., and N. Shephard. 1993. Structural time series models. In Handbook of statistics, ed. G. S. Maddala, C. R.
Rao and H. D. Vinod, Vol. 11. Barking: Elsevier Science Publishers.
Harvey, A. C. 1997. Trends, cycles and autoregressions. The Economic Journal 107 (440):192–201. doi:10.1111/1468-
0297.00152.
Harvey, A. C., S. G. B. Henry, S. Peters, and S. Wren-Lewis. 1986. Stochastic trends in dynamic regression models: An
application to the employment–output equation. Economic Journal 96 (384):975–85. doi:10.2307/2233168.
Harvey, A. C., and S. J. Koopman. 1992. Diagnostic checking of unobserved-components time series models. Journal of
Business & Economic Statistics 10 (4):377–89.
22 D. GUVEN ET AL.
Harvey, A. C., and S. J. Koopman. 1993. Forecasting hourly electricity demand using time varying splines. Journal of
American Statistical Association 88 (424):1228–36. doi:10.1080/01621459.1993.10476402.
Hendry, D., and K. Juselius. 2000. Explaining cointegration analysis: Part I. The Energy Journal 21 (1):1–42. doi:10.5547/
0195-6574-EJ-Vol21-No1-1.
Holtedahl, P., and F. L. Joutz. 2004. Residential electricity demand in Taiwan. Energy Economics 26 (2):201–04.
doi:10.1016/j.eneco.2003.11.001.
Hunt, L. C., G. Judge, and Y. Ninomiya. 2003a. Modeling underlying energy demand trends. In In energy in competitive
market: Essays in honour of colin robinson, ed. L. C. Hunt, 140–74. UK: Edward Elgar.
Hunt, L. C., G. Judge, and Y. Ninomiya. 2003a. Underlying trends and seasonality in UK energy demand: A sectoral
analysis. Energy Economics 25 (1):93–118. doi:doi:10.1016/S0140-9883(02)00072-5
Hunt, L. C., and Y. Ninomiya. 2003. Unravelling trends and seasonality: A structural time series analysis of transport oil
demand in the UK and Japan. The Energy Journal 24 (3):63–96. doi:10.5547/0195-6574-EJ-Vol24-No3-3.
Hunt, L. C., and Y. Ninomiya. 2005. Primary energy demand in Japan: An empirical analysis of long-term trends and
future CO2 emissions. Energy Policy 33 (11):1409–24. doi:10.1016/j.enpol.2003.12.019.
IEA (International Energy Agency). 2018a. The future of cooling report. https://doi.org/10.1787/9789264301993-en
IEA (International Energy Agency). 2018b. World energy balances 2018 report. https://doi.org/10.1787/data-00512-en
IEA (International Energy Agency). 2019a. Real residential electricity prices data. https://www.iea.org/statistics/prices/.
IEA (International Energy Agency). 2019b. Electricity statistics- Heat generation by source: Turkey. https://www.iea.org/
countries/turkey.
IMF (International Monetary Fund). 2018. World economic outlook, April 2018: Cyclical upswing, structural change.
Washington, DC: International Monetary Fund Publication Services.
Invidiata, A., and E. Ghisi. 2016. Impact of climate change on heating and cooling energy demand in houses in Brazil.
Energy and Buildings 130:20–32. doi:10.1016/j.enbuild.2016.07.067.
IPCC (Intergovernmental Panel on Climate Change). 2013. Summary for policymakers. In Climate change 2013: The
physical science basis. contribution of working group i to the fifth assessment report of the intergovernmental panel on
climate change, ed. T. F. Stocker, D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex,
and P. M. Midgley, 3–27, United Kingdom and New York, NY, USA: Cambridge University Press.
Javid, M., and A. Qayyum. 2014. Electricity consumption-GDP nexus in Pakistan: A structural time series analysis.
Energy 64:811–17. doi:10.1016/j.energy.2013.10.051.
Kilic, M., and E. Ozdemir. 2018. Long-term Energy Demand and Supply Projections and Evaluations for Turkey. In
Exergetic, Energetic and Environmental Dimensions, ed. I. Dincer, C. O. Colpan and O. Kizilkan, 115-132. Academic
Press.
Koopman, S. J., A. C. Harvey, J. A. Doornik, and N. Shephard. 2007. Stamp: Structural time series analyser, modeller and
predictor: STAMP 8. London: Timberlake Consultants Press.
Maddala, G. S., and I. Kim. 1998. Unit roots, cointegration, and structural change. Cambridge: University Press. ISBN:
978–0–521–58257–5.
NASA (National Aeronautics and Space Administration). 2015. NASA earth exchange global daily downscaled projec
tions (NEX-GDDP). http://dx.doi.org/10.7292/W0MW2F2G.
Ozcan, M. 2018. The role of renewables in increasing Turkey’s self-sufficiency in electrical energy. Renewable and
Sustainable Energy Reviews 82:2629–39. doi:10.1016/j.rser.2017.09.111.
Ozturk, H. K., H. Ceylan, O. E. Canyurt, and A. Hepbasli. 2005. electricity estimation using genetic algorithm approach:
A case study of Turkey. Energy 30 (7):1003–12. doi:10.1016/j.energy.2004.08.008.
Perez-Andreu, V., C. Aparicio-Fernandez, A. Martinez-Ibernon, and J. Vivancos. 2018. Impact of climate change on
heating and cooling energy demand in a residential building in a Mediterranean climate. Energy 165:63–74.
doi:10.1016/j.energy.2018.09.015.
Ricke, K., L. Drouet, K. Caldeira, and M. Tavoni. 2018. Country-level social cost of carbon. Nature and Climate Change 8
(10):895–900. doi:10.1038/s41558-018-0282-y.
Rodrigues, N., L. Losekann, and G. S. Filho. 2018. Demand of automotive fuels in Brazil: Underlying energy demand
trend and asymmetric price response. Energy Economics 74:644–55. doi:10.1016/j.eneco.2018.07.005.
Sa’ad, S. 2011. Underlying energy demand trends in South Korean and Indonesian aggregate whole economy and
residential sectors. Energy Policy 39 (1):40–46. doi:10.1016/j.enpol.2010.09.001.
Salvarli, H., and M. S. Salvarli. 2017. Trends on energy policy and sustainable development in Turkey. Energy Sources,
Part B: Economics, Planning, and Policy 12 (6):512–18. doi:10.1080/15567249.2016.1217284.
Scripps Institution of Oceanography. 2019. Scripps CO2 Program: Carbon dioxide concentration at Manua Loa
observatory. Accessed September 23, 2019. https://scripps.ucsd.edu/programs/keelingcurve/.
Shi, G., X. Zheng, and F. Song. 2012. Estimating elasticity for residential electricity demand in China. The Scientific
World Journal 2012:1–6. doi:10.1100/2012/395629.
Sozen, A., M. A. Akcayol, and E. Arcaklioglu. 2006. Forecasting net energy consumption using artificial neural network.
Energy Sources, Part B: Economics, Planning, and Policy 1 (2):147–55. doi:10.1080/009083190881562.
Sun, Y. 2015. Electricity prices, income and residential electricity consumption. In The Dynamic Energy Landscape, 33rd
USAEE/IAEE North American Conference, October 25–28. Pittsburgh, PA, USA.
TEDC (Turkish Electricity Distribution Corporation). 2018. Electricity distribution and consumption statistics of
Turkey report [in Turkish].
TurkStat (Turkish Statistical Institute). 2018a. Population by years, age group and sex, 1935–2018 Report. Accessed
August 11, 2019. http://www.tuik.gov.tr/PreIstatistikTablo.do?istab_id=1588.
TurkStat (Turkish Statistical Institute). 2018b. Population of province/district centers and towns/villages by province
and sex, population density by province, 2007–2018 report.
TurkStat (Turkish Statistical Institute). 2019a. Electricity generation and shares by energy resources. Accessed October 2,
2020. http://www.tuik.gov.tr/PreIstatistikTablo.do?istab_id=1578.
TurkStat (Turkish Statistical Institute). 2019b. Population of provinces by years report. Accessed August 11, 2019. http://
www.tuik.gov.tr/PreIstatistikTablo.do?istab_id=1590.
Ugur, F., and N. Gençyılmaz. 1990. Türkiye uzun dönem elektrik enerjisi üretim-yatırım planlama çalışması [A
long term electrical energy production-investment plan in Turkey] Fifth Energy Congress of Turkey, Ankara,
Turkey.
UNFCCC. 2015. Republic of Turkey intended nationally determined contribution report. Accessed September 23, 2019.
https://www4.unfccc.int/sites/submissions/INDC/Published%20Documents/Turkey/1/The_INDC_of_TURKEY_v.
15.19.30.pdf.
Uzlu, E. 2019. Application of Jaya algorithm-trained artificial neural networks for prediction of energy use in the nation
of Turkey. Energy Sources, Part B: Economics, Planning, and Policy 14 (5):183–200. doi:10.1080/
15567249.2019.1653405.
Welsch, H. 1989. The reliability of aggregate energy demand functions: An application of statistical specification error
tests. Energy Economics 11 (4):285–92. doi:10.1016/0140-9883(89)90044-3.
World Bank. 2018a. World development indicators. GDP per capita- Turkey. Accessed September 23, 2019. https://data.
worldbank.org/indicator/NY.GDP.PCAP.CD?locations=TR.
World Bank. 2018c. World development indicators. manufacturing value added- Turkey. Accessed September 23, 2019.
https://data.worldbank.org/indicator/NV.IND.MANF.KD?locations=TR.
World Bank. World development indicators. Turkey: Urban Development. Accessed September 23, 2019. https://data.
worldbank.org/topic/urban-development?locations=TR.

Impact of Climate Change On Sectoral Electricity Demand in Turkey

Uploaded by

Copyright:

Available Formats

Impact of Climate Change On Sectoral Electricity Demand in Turkey

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Impact of Climate Change On Sectoral Electricity Demand in Turkey

Uploaded by

Copyright:

Available Formats

Energy Sources, Part B: Economics, Planning, and Policy

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/uesb20

Impact of climate change on sectoral electricity

Denizhan Guven, M. Ozgur. Kayalica, Gulgun. Kayakutlu & Erkan. Isikli

To link to this article: https://doi.org/10.1080/15567249.2021.1883772

Published online: 10 Mar 2021.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

Impact of climate change on sectoral electricity demand in Turkey

Wind Biogas/Biomass Thermal Hydro Solar

Share of total final energy

Figure 2. Energy use for cooling in buildings (Source: IEA, 2018a).

2.1. Previous electricity demand modeling for Turkey

2.2. Literature review of the modeling approach

2.3. Effects of climate on electricity demand modeling

Table 2. Some selected STSM studies.

4. Data and estimation

4.2. Estimation results

4.2.1. Residential electricity

Table 3. Brief statistical analysis of variables.

ert ¼ 0:3766ert 1 þ 0:1256yt þ 0:1019yt 1 0:1263prt 0:197URt þ 0:1898URt 1 þ 5:3971CDDt

4.2.2. Industrial electricity

4.2.3. Commercial electricity

4.3. Forecasting assumptions

5. Results and discussion

Table 5. Comparison of price and income elasticities.

Region/ Residential Industrial Commercial

Figure 10. Vicious circle of space cooling demand.

You might also like