Contents lists available at ScienceDirect
The Electricity Journal
journal homepage: www.elsevier.com/locate/tej
Electricity demand modeling in Saudi Arabia: Do regional differences
matter?
Jeyhun I. Mikayilova,b,*, Fakhri J. Hasanova,c, Waheed Olagunjud, Mohammad H. Al-Shehrie
a King Abdullah Petroleum Studies and Research Center, PO Box 88550, Riyadh, 11672, Saudi Arabia
b Azerbaijan State University of Economics (UNEC), Istiqlaliyyat Str., 6, Baku, Azerbaijan
c The George Washington University, 2115 G Street, NW, Washington, DC, 20052, USA
d Ministry of Energy, King Saud St., Al Wizarat, Riyadh, 12622, Saudi Arabia
e Saudi Electricity Company, Takhassusi St., Riyadh, 12333, Saudi Arabia
A R T I C L E I N F O
Keywords:
Electricity demand
Price elasticity
Cointegration
Regions
Saudi Arabia
A B S T R A C T
The paper examines the electricity demand behavior for Saudi Arabia, using annual data for the period of
1990–2016, at regional level. The study finds that income, price and population are the main drivers of elec-
tricity demand at regional level. Although, the impacts vary across regions (central, eastern, southern and
western), the estimated elasticities all are statistically significant, in both long and short run, and have the
expected signs for all the regions. The income, price and population elasticities range from 0.10 to 0.93, from
-0.63 to -0.06, and from 0.24 to 0.95, respectively, across regions in the long run. In the short run these intervals
are (0.05, 0.47), (−0.27, −0.01) and (0.13, 1.49) respectively for income, price and population across the
regions. The findings reported in this paper, should assist policy makers to develop insights about the potential
regional impact of changes to electricity prices, income and population patterns.
1. Introduction
Since energy is a pervasive input to all business and recreational
activities, total energy demand is an important indicator that helps
explain the pattern of economic development within a country.
Identifying and understanding the key determinants of electricity de-
mand is therefore important for the economic prosperity of a country,
since the availability of reliable electricity directly affects the prospects
of sustainable economic development. With important mega projects
already in the works and national transformation programs such as the
National Industrial Development and Logistics Program (NIDLP, 2019)
already being implemented, understanding the existing and projected
behaviour of electricity demand is more important now than ever be-
fore.
While it is relevant to conduct a study that investigates the de-
terminants of aggregate electricity demand in KSA, this is a topic that
has already been addressed by the existing literature to some extent
(Atalla and Hunt, 2016, inter alia). However, aggregate level electricity
demand has not been studied at regional level. Hence, there is a need
for a research to investigate the aggregate level electricity demand
...
Z Score,T Score, Percential Rank and Box Plot Graph
Contents lists available at ScienceDirectThe Electricity J
1. Contents lists available at ScienceDirect
The Electricity Journal
journal homepage: www.elsevier.com/locate/tej
Electricity demand modeling in Saudi Arabia: Do regional
differences
matter?
Jeyhun I. Mikayilova,b,*, Fakhri J. Hasanova,c, Waheed
Olagunjud, Mohammad H. Al-Shehrie
a King Abdullah Petroleum Studies and Research Center, PO
Box 88550, Riyadh, 11672, Saudi Arabia
b Azerbaijan State University of Economics (UNEC),
Istiqlaliyyat Str., 6, Baku, Azerbaijan
c The George Washington University, 2115 G Street, NW,
Washington, DC, 20052, USA
d Ministry of Energy, King Saud St., Al Wizarat, Riyadh,
12622, Saudi Arabia
e Saudi Electricity Company, Takhassusi St., Riyadh, 12333,
Saudi Arabia
A R T I C L E I N F O
Keywords:
Electricity demand
Price elasticity
Cointegration
Regions
Saudi Arabia
2. A B S T R A C T
The paper examines the electricity demand behavior for Saudi
Arabia, using annual data for the period of
1990–2016, at regional level. The study finds that income, price
and population are the main drivers of elec-
tricity demand at regional level. Although, the impacts vary
across regions (central, eastern, southern and
western), the estimated elasticities all are statistically
significant, in both long and short run, and have the
expected signs for all the regions. The income, price and
population elasticities range from 0.10 to 0.93, from
-0.63 to -0.06, and from 0.24 to 0.95, respectively, across
regions in the long run. In the short run these intervals
are (0.05, 0.47), (−0.27, −0.01) and (0.13, 1.49) respectively for
income, price and population across the
regions. The findings reported in this paper, should assist policy
makers to develop insights about the potential
regional impact of changes to electricity prices, income and
population patterns.
1. Introduction
Since energy is a pervasive input to all business and
recreational
activities, total energy demand is an important indicator that
helps
explain the pattern of economic development within a country.
Identifying and understanding the key determinants of
electricity de-
mand is therefore important for the economic prosperity of a
country,
since the availability of reliable electricity directly affects the
prospects
of sustainable economic development. With important mega
3. projects
already in the works and national transformation programs such
as the
National Industrial Development and Logistics Program
(NIDLP, 2019)
already being implemented, understanding the existing and
projected
behaviour of electricity demand is more important now than
ever be-
fore.
While it is relevant to conduct a study that investigates the de-
terminants of aggregate electricity demand in KSA, this is a
topic that
has already been addressed by the existing literature to some
extent
(Atalla and Hunt, 2016, inter alia). However, aggregate level
electricity
demand has not been studied at regional level. Hence, there is a
need
for a research to investigate the aggregate level electricity
demand
accross regions.
The objective of this research is to understand how determinants
of
the electricity consumption shape it over time in the Central,
Eastern,
Southern and Western regions of Saudi Arabia and suggest
policy in-
sights accordingly.
Note that this 4-region classification is adopted by the Saudi
Electricity Company (SEC) and sum of these four regions’
consumption
4. add up to the total electricity consumption of the Kingdom.
The following points motivated us to conduct electricity
consump-
tion analysis at the regional level. First, the regional dimension
is im-
portant because of the differences in weather patterns (and
subse-
quently the electricity demand profiles) across the country.
Second, the
distribution of residential, commercial, and industrial activities
is dif-
ferent across the regions and this may imply different
relationships
between electricity consumption and its drivers. For example, in
the
eastern region the electricity consumption is mainly industry-
driven
whereas in western region, where we have the holy cities of
Makkah
and Madinah, this is mostly population/residential driven
(SAMA,
2019). Third, the implementation of the mega projects
mentioned
above will also have implications for electricity demand at the
regional
level. Finally, recent analysis show that different regions react
differ-
ently to the energy, of which electricity price reform, one of the
key
initiatives in the Fiscal Balance Program of the Saudi Vision
2030
(Alyamani et al., 2019).
https://doi.org/10.1016/j.tej.2020.106772
6. regions
of Saudi Arabia over the period 1980−1992. However, he did a
panel
analysis and did not estimate region-specific
parameters/elasticities,
and hence is different from time series that we conduct in this
research.
Additionally, he did not address integration-cointegration and
other
properties of the data used such as cross-sectional dependency,
which
can lead a serious issue such as bias and inconsistencies and
thus mis-
leading policy recommendations. Second, it investigates the
demand for
electricity at a disaggregated regional level, which takes into
account
the region specific features of electricity demand behavior;
Third, it
takes into account the impact of demographic factors which may
play a
significant role in electricity demand formation; Fourth, it uses
the
more recent data, which partially enables us to see the impact of
the
ongoing energy price reforms and consequences of the low oil
price
environment on electricity demand.
From a policy perspective, given the current transition from a
heavily subsidized electricity price environment to a market-
based
price environment where subsidies are gradually phased out, it
is im-
portant to consider the impact of this transition from a regional
per-
7. spective. The findings reported in this study would help a better
un-
derstanding of the regional impact (on electricity demand) of
different
price policy scenarios and changes in population and income.
The
findings will also be useful in determining which regions, and
how
much, to provide transitional support to in order to alleviate
some of
the financial hardships associated with rising electricity prices.1
This paper employs cointegration and equilibrium error
correction
(ECM) methodology to develop long- and short-run price,
income and
demographic elasticities for regional electricity demand.
The study concludes that income, price and population are the
main
drivers of electricity demand at regional level as theoretically
expected.
Although, the impacts vary across regions, the estimated
elasticities all
are statistically significant, in both long and short run, and have
the
expected signs for all the regions. The income, price and
population
elasticities range from 0.10 to 0.93, from -0.63 to -0.06, and
from 0.24
to 0.95, respectively, across regions in the long run. In the short
run
these intervals are (0.05, 0.47), (-0.27, -0.01) and (0.13, 1.49)
respec-
tively for income, price and population across the regions. The
obtained
8. Speed of Adjustment (SoA) coefficients are significant in all
cases in-
dicating the short-run deviations from the long-run relationship
con-
verge back to the equilibrium path.
The rest of the paper is structured as follows; Section 2 reviews
the
literature on electricity demand modeling in the case of Saudi
Arabia,
Section 3 contains the theoretical framework and Section 4
briefly de-
scribes the methodology used, Section 5 presents the data.
Estimation
results are presented in Section 6. In Section 7, we discuss our
key
findings, while Section 8 concludes the study and provides the
policy
implications where we link our findings to the current policy
environ-
ment.
2. Literature review
In this section, we highlight some of the key trends in the
literature
on residential, industrial and total electricity consumption in
Saudi
Arabia given that there are only two studies, to the best of our
knowledge, analyzing the regional electricity consumption in
Saudi
Arabia. A recent study by Alyamani et al. (2019) discussed
regional
aspects of electricity consumption in residential sector.
However,
9. Alyamani et al. (2019) is descriptive-type study and does not
estimate
elasticities. Diabi (1998) studied total electricity consumpti on
for the
five regions of Saudi Arabia, employing the data spanning from
1980 to
1992. He estimated the kingdom-wise elasticities not regional-
specific
ones. The study is relatively old which might not reflect the
electricity
demand behaviour for the current period, due to the substantial
changes in terms of economic development. In addition, Diabi
(1998)
uses panel estimation techniques, without performing
integration-co-
integration analyses, did not address the potential cross-
sectional de-
pendency within regions. The mentioned limitations might
result in
biased estimation results and consequently misleading
conclusions.
The studies investigating electricity consumption employed a
wide
range of analytical techniques that are not necessarily
quantitative or
econometrically sound. We found studies that focused on
qualitative
analysis, Granger causality analysis, simulations based on
optimization
models, and a range of econometric estimation methods.
For example, both Hagihara (2013) (who simply described the
en-
ergy outlook in KSA) and Alrashed and Asif (2014) who
conducted a
10. survey analysis of residential electricity consumption (REC) in
the
Eastern province of KSA can be seen as qualitative analysis.
Matar (2017) and Matar and Anwer (2017) are examples of
studies
that adopted a simulation-based approach when investigating
elec-
tricity consumption. Both studies explored the impact of
electricity
price changes in KSA on residential electricity consumption in
2011 and
2015, respectively. In both studies, the simulations were
conducted
using an optimization based partial equilibrium model. They did
not
explicitly consider income and demographic effects and there
were no
elasticities reported due to the nature of the study.
Since the focus of this review is on econometric studies in line
with
the nature of our research here, we continue by reviewing the
existing
literature with emphasis on the type of data that was used, the
econometric methodology and the specifications employed, the
em-
pirical analysis strategy adopted (e.g., whether a study at hand
ad-
dressed stochastic properties of the data) for the study.
We compare the model specifications to the standard
specification
as dictated by the theories, which requires that we account for
income,
price and demographic effects (Beenstock and Dalziel, 1986;
11. Liddle and
Lung, 2010; Hasanov, 2019, inter alia). This is important
because the
results from studies that do not account for all factors could
potentially
contain some omitted variable bias.
For industrial electricity demand in KSA, there are some studies
investigating this relationship. Al-sahlawi (1999) utilized
aggregate
time series data and Eltony and Mohammad (1993) and Liddle
and
Lung (2010) utilized country-level panel data but did not report
KSA
specific estimates. In their specifications, Al-sahlawi (1999)
considered
only income, Eltony and Mohammad (1993) considered income
and
price, and Liddle and Lung (2010) considered only urbanization
rates.
More recently, Hasanov (2019) studied the determinants of
industrial
electricity demand for Saudi Arabia by analyzing the core
drivers of
electricity consumption in industrial sector.
As mentioned earlier, the results from studies that do not
account
for income, price and demographic effects might be biased.
Since first
three studies mentioned above do not account for all 3 factors,
they
may contain some bias. Furthermore, Al-sahlawi (1999) and
Eltony and
Mohammad (1993) did not consider the integration-
cointegration
12. properties of the variables included in their analysis before
using OLS
estimation. Therefore, their results might be biased from the
spurious
regression perspective. In addition, both studies are quite old
and the
relationship between the variables of interest might demonstrate
new
path with the recent data.
Next, we turn our attention to key studies on residential
electricity
demand. In this area, there were some studies with a regional
focus.
1 Ideally, a regional analysis for each electricity consumer
category would
provide further insights about the impact of rising electricity
prices on different
consumer categories and allow for a more targeted support
approach, but that is
beyond the scope of our analysis. Authors currently are working
on this task.
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772
2
Precisely speaking, they focused on the Gulf Council Countries
(GCC) as
a region and not the regions within KSA as we do in this study.
The
earliest study we found was the work of Eltony and Mohammad
(1993)
13. who examined residential electricity demand for a panel of GCC
countries including KSA and found long run (LR) income and
price
elasticities for GCC countries of 0.20 and −0.14, respectively.
The
study did not explicitly account for demographic effects. Al -
Sahlawi
(1999) estimated the short run (long run) income and price
elasticity of
residential electricity demand as 0.13 (0.70) and −0.10 (−0.50),
re-
spectively. This study also did not explicitly account for
demographic
effects. Both studies used OLS and did not account for the
integration-
cointegration properties of the variables.
Atalla and Hunt (2016) use a structural time series model
(STSM)
and data from 1985 to 2012 to investigate residential electricity
con-
sumption for GCC countries. Unlike other studies examined
thus far,
they accounted for all three factors required to explain
residential
electricity consumption. They also used Cooling and Heating
Degree
Days variables as a proxy for weather conditions. For KSA, they
found
the following long-run elasticities: 0.48 for income -0.16 for
price and
0.80 for population, which was the variable representing the
demo-
graphic effect. The study also concludes that the short-run (SR)
price
and population elasticities are-0.16 and 4.20, respectively,
14. while in-
come does not affect the demand in the short-run.
With regards to modelling total electricity consumption, there
are
several studies for KSA.
Al-Faris (2002) used a Vector Error Correction Model (VECM)
ap-
proach and data from 1970 to 1997 to find LR (SR) income and
price
elasticities for KSA of 0.05 (1.65) and -0.04 (-1.24),
respectively.
Narayan and Smyth (2009) using data from 1974 to 2002 and
FMOLS
method reported long run income elasticity of electricity
consumption
for KSA of -3.07. In addition, the paper does not
explain/interpret the
found unusual negative and substantially higher income
elasticity.
Liddle and Lung (2010); Karanfil and Li (2015) and
Mohammadi
and Amin (2015) use ECM and panel data for many countries
including
KSA to investigate the causal relationship between total
electricity
consumption and urbanization. They found that GDP per capita
and
urbanization granger cause total electricity consumption per
capita.
Hasanov et al. (2017) use an equilibrium correction model and
panel data for a number of oil exporting countries including
KSA to
15. investigate the relationship between GDP, residential electricity
con-
sumption, foreign direct investment and employment. One of
their key
findings was that employment Granger-causes residential
electricity
consumption in the short run. Similarly, Salahuddin et al.
(2015) using
panel data for GCC countries found that GDP per capita granger
causes
total electricity consumption per capita and reported a LR
income
elasticity of 0.41.
Diabi (1998) analyzed regional total electricity consumption in
KSA,
based on panel data (1980–1992) for five KSA regions at the
time (CR,
WR, ER, SR and NR). The study compared the results of
different esti-
mation methods (OLS, CHTA, CCTA, FE, RE and MLE) and
accounted
for income, price and demographic effects using the
urbanization rate.
He reported long run elasticities for KSA of 0.09 to 0.49 for
income,
−0.14 to 0.00 for price and 0.93–1.30 for urbanization.
Corresponding
SR elasticities are income (0.05 to 0.33), price (-0.12 to 0.00)
and ur-
banization (0.62–1.10).
In summary, there are a number of studies examining
residential,
industrial and total electricity consumption in KSA, but to the
best of
16. our survey, there are no studies that examine regional total
electricity
consumption. Considering this fact, the current study aims to in-
vestigate the determinants of electricity demand for Saudi
Arabia at the
regional level using different cointegration techniques.
3. Theoretical framework
We use a standard formulation suggested by the theories such as
demand-side approach :
Electricity use = F (price, income, population)
Where Electricity use is the total regional electricity demand,
Income is
income proxy, Population is population size of the appropriate
region
and Price is real electricity price.
The econometric functional relationship can be formulated as
follow:
= + + + +Electricity use α α Income α Population α Price e0 1 2
3
Since increases in income and population increase electricity
de-
mand, while price increases negatively affect the demand of
electricity,
the expected signs for the coefficients α α, and 1 2 are positive
while α3 it
expected to be negative. e
All the variables are in logarithmic form in above specification;
hence, the coefficients are elasticities, which capture the
17. percentage
change in electricity use as a result of a 1% change in the
variable
considered.
4. Methodology
For estimating the long-run relationships between the variables
of
interest, the paper uses cointegration techniques such as
Dynamic
Ordinary Least Squares (DOLS), Canonical Cointegration
Regression
(CCR) and FMOLS. For estimating the short-run elasticities and
Speed of
Adjustment (SoA), we used ECM in the framework of General
to
Specific Modeling Strategy (Gets) (Campos et al., 2005; Hendry
et al.,
2008; Doornik and Hendry, 2009; Doornik, 2009; Doornik and
Hendry,
2018, inter alia). In addition, since we are using time series data
the
variables should be tested for their integration properties, and
the
Dickey and Fuller (1981) unit root test is used for this exercise.
For
cointegration exercises, we used the Engle and Granger (1987)
coin-
tegration test. Considering that, all the aforementioned
econometric
methods and tests are widely used and well known to the
researchers,
we are not describing them here. Interested readers can refer to
Saikkonen (1992); and Stock and Watson (1993) for DOLS, and
Phillips
18. and Hansen (1990) for FMOLS, and Park (1992) for CCR
methods.
5. Data
In this section, we describe the data that was used in our
analysis by
discussing some descriptive statistics and technical properties
of the
data.
The paper uses annual time series data for the sample spanning
from
1990 to 2016, which is chosen based on the data availability.
The data
used in the study are described below:
Electricity use represents regional electricity consumption in
each
of the four regions measured in Mwh; these data are taken from
the
Saudi Electricity Company (SEC).
Price is the regional (weighted based on the regional
consumption
types) real weighted average electricity price, in SAR per TOE
(Saudi
Arabian Riyals per Ton of Oil Equivalent). To convert the
nominal
electricity prices to real values Consumer Price Index (CPI,
index
2010 = 100) and GDP deflator (PGDP, 2010 = 100) values are
used2 .
CPI data is taken from Saudi Arabia Monetary Authority’s
(SAMA,
2017) website (SAMA annual stats: Cost of Living Index by
19. Divisions;
http://www.sama.gov.sa/en-US/EconomicReports/Pages/
YearlyStatistics.aspx). PGDP data taken from the web page of
General
Authority for Statistics of Saudi Arabia
(https://www.stats.gov.sa/en/
823). Nominal electricity price is the nationwide or aggregate
price as
Saudi Arabia does not apply different prices to different
regions.
2 The nominal price values were converted to real prices using
GDP deflators
for the Central and Southern regions, and CPI for the Eastern
and Western re-
gions depending on which deflated price produce economically
meaningful and
statistically significant results.
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772
3
http://www.sama.gov.sa/en-
US/EconomicReports/Pages/YearlyStatistics.aspx
http://www.sama.gov.sa/en-
US/EconomicReports/Pages/YearlyStatistics.aspx
https://www.stats.gov.sa/en/823
https://www.stats.gov.sa/en/823
Income is proxied by GDP and disposable income (DI), both in
Millions SAR, at 2010 prices. GDP data is taken from the web
page of
General Authority for Statistics of Saudi Arabia (General
Authority for
20. Statistics (GaStat), 2018). DI data is calculated by KAPSARC
re-
searchers.
Population is regional population, in persons, as a proxy for the
demographic factors. This data was aggregated from the General
Authority for Statistics of Saudi Arabia as it is available for 13
provinces
(SAMA, 2019).
Fig. 1 demonstrates the graphs of the shares of regional
electricity
consumption in total country level electricity consumption.
As can be seen from the figure the share of electricity
consumption
significantly higher in the eastern region in early 90th, while
decreasing
over time it approached to the consumption levels in central and
wes-
tern region in recent years.
Theoretically, the long-run analysis can be conducted in logs or
levels; here we choose the log-log specification due to its ease
of in-
terpretation and compatibility with existing literature (to allow
easier
comparisons with previous and future studies).
6. Empirical estimation results
6.1. Unit root test results
As is standard practice in the literature, the main concern lies in
dealing with stochastic properties of the variables and hence the
Augmented Dickey-Fuller (ADF) Unit Root test was employed
21. for each
variable in levels and first differences. In unit-root testing, the
max-
imum lag length is taken to be two and optimal lag is chosen
based on
Schwarz information criterion (SIC). Table 1 displays the
results of the
Unit Root tests.
Briefly, based on the results displayed in Table 1 we can
conclude
that all variables are integrated of the first order [I(1)] and there
is
potential for cointegration relationship. Hence, we can proceed
to the
next step and test variables for the common trend. The next
section
provides the results of cointegration analyses.
6.2. Cointegration test results
To test the existence of cointegration relationship, i.e., whether
the
variables share the long-run common trend, we employed the
Engle-
Granger (EG) cointegration test and the results are given in
Table 2.
As the table demonstrates, for all regions except the EOA, we
found
conclusive proof of cointegration. For the three regions, all the
p-values
from the EG tests were below the 5% level. This confirmed the
presence
of a cointegration relationship at the 5% significance level. For
22. further
investigation of the existence of the long-run relationship in the
case of
the Eastern region, we employed the Variable Addition Test
(VAT) for
cointegration, proposed by Park (1990), which states existence
of co-
integration as a null hypothesis. The test statistic is 4.932 with
the p-
value of 0.085, concluding the cointegration relationship at 8.5
%
significance level. Considering the results of employed tests, we
con-
clude that, for all the regions there is a long-run relationship.
6.3. Long-run estimation results
After concluding the cointegration relationship among the
variables
the long-run estimation results can be interpreted so they are
not
spurious. The long-run estimation results are provided in Table
3. To
have an idea of the impacts in terms of magnitude ranges, we
reported
the 99 % confidence intervals for the estimated
coefficients/elasticities.
The interval estimator provides more information about the re-
presentativeness quality of the estimated coefficient, in addition
to
delivering the limits of the impact.
Based on the estimation results presented in Table 3, one can
see
that for all the regions, the impacts of income, price and
population are
23. economically meaningful as they take right signs and
statistically sig-
nificant at 1% level. As can be seen from Table 3, the income
elasticity
of electricity demand varies across regions, demonstrating more
stable
behavior in the Western region and wider ranged in the Central
region.
The highest income elasticity found in Central region, while it
has the
smallest value in the Eastern region. Overall, the income
elasticity
ranges from 0.102 to 0.931 across the regions.
Price elasticity of electricity demand is found to be ranging
from
-0.607 to -0.060 across the regions.
The impact of population demonstrates more similar path across
regions, having differences as well. Numerically the population
elasti-
city ranges from 0.243 to 0.947.
6.4. Short-run estimation results
For the short-run analysis, we applied Gets to ECM to
determine the
relationship for each region. We start with a general ECM
specification,
which includes the error correction term (ECT),
contemporaneous va-
lues of all independent variables and 2 lags of all the variables.
We then
exclude variables from the analysis based on the test proposed
by the
methodology and end up with the final short-run specification
24. (e.g., see
Campos et al., 2005 inter alia). We are not reporting the
General Un-
restricted Models here to save space, but they are available from
the
authors upon request. Table 4 documents the final ECM
specification
Fig. 1. Shares of regional electricity demands in total, %.
Table 1
ADF Unit Root Test results.
Region Variable (in logs) Level First Difference
National gdp −0.948 −4.269***
di −0.138 −5.147***
Central dele_coa −0.976 −4.962***
pele_coa −1.274 −4.514***
pop_coa −0.271 −5.618***
Eastern dele_eoa −1.576 −4.403***
pele_eoa −1.782 −4.465***
pop_eoa −0.582 −3.806***
Western dele_woa −0.903 −6.994***
pele_woa −1.898 −4.438***
pop_woa −0.139 −4.668***
Southern dele_soa −0.444 −5.110***
pele_soa −1.327 −4.534***
pop_soa −0.670 −4.476***
Note: dele = demand for electricity, COA = central operating
region,
25. EOA = eastern operating region, SOA = southern operating
region,
WOA = western operating region, pop = population, pele =
electricity price;
*** =Significant at the 1% level, intercept only case is chosen
based on the
analyses.
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772
4
and the post-estimation test results.
As the table demonstrates the short-run final specifications pass
all
the diagnostic and misspecification tests, hence the results are
inter-
pretable. Two observations are worth stating. First, all the
remained
regressors in the final ECM have economically meaningful and
statis-
tically significant impact on the electricity consumption in all
regions.
Second, this is true for the contemporaneous values of all the
drivers,
i.e., the income, price and population.
The estimated Speed of Adjustment (SOA) coefficients
(coefficient
of the ECT term) are statistically significant and negative in all
the
regions, confirming the existence of stable long-run relationship
among
26. the variables.
7. Discussion of the results
In this section, we present the key results and insights from our
analysis on regional electricity demand. The key results are
summarized
in Tables 3 and 4. Table 3 displays the estimated long run
elasticities for
each region and Table 4 displays the short run elasticities.
As implied by Table 3, the long-run income elasticities range
from
0.10 to 0.93 across the regions, the long-run price elasticities
range
from -0.61 to -0.06, and the long-run population elasticities
range from
0.24 to 0.95. These results are in line with our expectation of
positive
elasticities for income and population, and negative elasticities
for
prices that we discussed in the Theoretical Framework section.
In terms
of magnitudes, our long-run income elasticities are smaller than
finding
of Al-Faris (2002), which is the only study devoted to total
electricity
after 2000, and seem to be bigger (1.65) than the expected
magnitude.
In this regard, it makes sense to find income elasticity being
smaller
than his finding. When it comes to long-run price elasticity, the
found
range is in line with the expectation for the developing country
case.
For example, Atalla and Hunt (2016) estimated the long-run
27. price
elasticity for residential demand to be -0.16, which is well
contained in
our interval. In addition, since the previous studies investigated
the
electricity demand modeling at overall country level, our results
are not
directly comparable with their results. The country level
elasticities can
be seen as the average representative for the whole country,
while the
regional ones might take into account the region-specific
features.
With an income elasticity of 0.93 (the upper bound of the con-
fidence interval), electricity demand in the Central region is the
most
sensitive to changes in income. The Eastern region is the most
sensitive
to changes in population as the population elasticity of
electricity de-
mand is the highest at 0.95 (the upper bound of the confidence
in-
terval).
With a price elasticity of −0.61 (The lower bound of the
confidence
interval. The western is chosen with the highest impact since
the nar-
rower confidence interval in comparison with east), the Western
region
displays the largest sensitivity to changes in prices. Comparing
the sizes
of all the elasticities for each region, income has the largest
impact on
electricity demand in the Central region, population has the
28. largest
impact in the Eastern region, while price has highest impact in
the
Western region.
Comparing the elasticities across regions, we observe that in
abso-
lute terms, the price elasticity is lowest in the Southern region
(-0.13,
Table 2
Cointegration test results.
COA EOA SOA WOA
Value P-value Value P-value Value P-value Value P-value
EG tau-statistic −5.413 0.013** −3.598 0.226 −5.273 0.037**
−4.168 0.040**
EG z-statistic −561.841 0.000*** −18.136 0.189 −60.934
0.000*** −21.866 0.033**
Notes: p-values are MacKinnon (1996) probability values; **
and *** stand for the rejection of the null of no cointegration at
5% and 1% significance level,
respectively.
Table 3
Long-run Estimation results.
Regressor Region
COA EOA SOA WOA
Income* (0.403, 0.931) (0.102, 0.204) (0.120, 0.360) (0.426,
0.470)
29. Price (−0.580,
−0.362)
(−0.628,
−0.356)
(−0.132,
−0.060)
(−0.607,
−0.427)
Population (0.243, 0.661) (0.771, 0.947) (0.335, 0.645) (0.776,
0.840)
Notes: Dependent variable is dele_i and i takes COA, EOA,
SOA and WOA, re-
spectively; * for the Central and the Southern regions GDP used
as an income
measure; For the Eastern and Western regions Disposable
Income is used as
income proxy; Numbers in parentheses are 99 % confidence
intervals for the
obtained elasticities.
Table 4
Short-run Estimation results.
Independent Variables (DLOGs) Region
Central Eastern Western Southern
ECT (−0.293, −0.151) *** (−0.892, −0.532) *** (−0.984,
−0.746) *** (−1.318, −0.924) ***
income (0.152, 0.414) ** (0.054,0.182) * (0.075, 0.221) *
(0.159, 0.471) *
30. price (−0.079, −0.013) * (−0.234, −0.088) ** (−0.274, −0.150)
*** (−0.103, −0.029) *
population (0.128, 0.616) * (0.168, 0.412) ** (0.755, 1.485) ***
(0.338, 0.600) ***
dele (−1) (0.118, 0.386) *
income (−1) (−0.212, −0.078) **
price (−1)
population (−1) (−0.552, 0.134) (−1.399, −0.839) ***
TESTS (p-values)
Serial Correlation LM 0.344 0.062 0.067 0.400
Normality test (Jarque-Bera) 0.850 0.856 0.522 0.300
Heteroscedasticity (White) 0.636 0.870 0.484 0.873
Ramsey RESET Test 0.471 0.153 0.133 0.942
Notes: Dependent variable is DLOG(DELE_i) and i takes COA,
EOA, WOA and SOA; ***, **, * stand for rejecting a null
hypothesis at 1%, 5% level and 10 % significance
levels, respectively. All independent variables except ECT are
in DLOG.
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772
5
the lower bound of the confidence interval), the population
elasticity is
lower in the Central and Southern regions and the income
elasticity is
lower in the Eastern and Southern regions. Below we rationalize
these
findings based on regional characteristics.
We suspect that the low price elasticity in the Southern region
is a
31. result of the relatively lower levels of income in the south
(Household
Income and Expenditure Survey, GaStat, 2018). Due to the
relatively
lower income levels, electricity consumption (which is mainly,
about
62%, residential consumption in the region) is already
conservative and
optimized, and there are not many avenues to further reduce
con-
sumption as a result of price increases. This view is reinfor ced
by the
relatively low income elasticity for the southern region.
In order to explain the regional population elasticities, it is im-
portant to highlight that population trends are explained by
migration
patterns, births and deaths. In the Southern region where
emigration is
prevalent due to the search for better economic opportunities in
other
regions by adults and population increases are mainly driven by
births,
a lower population elasticity is potentially explained by
population
growth being driven by mainly births. In comparison to adults,
young
children tend to consume less electricity, hence the lower
population
elasticity in the south is potentially due to population growth
being
driven by births. The Southern and Eastern regions having lower
in-
come elasticities is potentially explained by the economic
character-
istics of the regions. In the Eastern region, which is heavily in-
32. dustrialized, the income elasticity being low makes sense
because of the
relatively higher levels of income in the region. Changes in
income are
less likely to impact the level of electricity consumption when
income
levels are high (Chang et al., 2016, inter alia). Not surprisingly,
in the
Southern region where income levels are lower, the income
elasticity is
also low. This is potentially explained by the conservative
approach to
electricity consumption (of which about 62% is resudential
consump-
tion in the Southern region) at relatively lower income levels.
This
makes consumers less sensitive to income changes. As Chang
(1977,
1980) and Chang and Hsing (1991) discusses, demand for a
particular
thing might be luxury for some level of income. At that level of
income,
depending its level, the income elasticity starts to grow rapidly,
and
even might become bigger than unity. After a certain level of
income,
the additional increase in income does not contribute its
consumption
at the same rate. Hence, the income elasticity with respect to
that item
reduces and further increase does not change it significantly.
Therefore,
for some low level and for higher levels of income the income
elasti-
cities are expected to be low. While, it can be closer, even
bigger than
33. unity for income levels between the low and high ones. This
point is
also concluded by Chang and Hsing (1991) for residential
electricity
consumption in US, and by Chang et al. (2016) for electricity
con-
sumption in case of panel of countries. In this regard, relatively
low
income elasticities in cases of the Eastern and Southern regions
can be
explained by the above-mentioned points.
8. Conclusions and policy implications
In this study, we examined the impact of price, income and
popu-
lation (proxy for demographic effects) on electricity demand for
four,
central, eastern, southern and western, regions in Saudi Arabia.
To the
best of our knowledge, this is the first time series study to
examine the
determinants of regional electricity demand in Saudi Arabia.
The empirical results show the estimates of the regional long-
run
price elasticities to be around −0.5 in three regions and −0.1 in
the
Southern region. This leads to the conclusion that changes in
the price
of electricity are an effective market signal that has the
potential to
impact the demand for electricity.
From a policy perspective, based on the regional differences in
elasticities, policy makers can develop insights about the
34. potential re-
gional impact of changes in electricity prices. Furthermore,
based on
the estimated regional impact, policy makers are better
positioned to
strike a balance between the desire to encourage efficient
electricity
consumption (by transitioning to a market-based pricing scheme
with
higher prices) and the negative impact of the price increase.
Striking a
balance will require policy makers to determine the appropriate
level of
household and industrial support required to dampen the
negative
impact of price changes.
With long run elasticities that range between 0.24 and 0.95, the
strong impact of population on regional electricity demand
requires
governments to pay special attention to policies that impact the
total
and regional distribution of population, which in turn affect the
ag-
gregate and regional distribution of electricity demand. This is
espe-
cially relevant for the effective planning of generation,
transmission
and distribution networks across the country. The recently im-
plemented expat levy is a relevant example of a policy that has
the
potential to change the existing population dynamics.
Lastly, our results show that an increase (decrease) in income
(gross
35. domestic product and disposable income) leads to an increase
(de-
crease) in electricity consumption. This result proves that
electricity is a
normal good from an income perspective. It is important to note
that
there are big differences in the income elasticities across
regions. The
differences range from approximately 0.7 in the COA to 0.2 in
the EOA.
This implies that it is important to account for the regional
income
differences when designing policy for the electricity sector.
In summary, our findings constitute evidence on the
characteristics
of the regional determinants of electricity demand in Saudi
Arabia. The
policy implications discussed above represent our contribution
to the
discussion in the current policy environment in Saudi Arabia.
There is a room for future research to analyze the regional elec-
tricity demand based on types of consumers. This would enable
pol-
icymakers to see the main types by regions driving the demand
for
electricity and clearer picture of demand behavior.
Author contributions
All authors contributed equally to all aspects of the research re -
ported in this paper.
Declaration of Competing Interest
36. The authors declare that they have no known competing
financial
interests or personal relationships that could have appeared to
influ-
ence the work reported in this paper.
Acknowledgments
The views expressed in this study are those of the authors and
do not
necessarily represent the views of their affiliated institutions.
We are
responsible for all error and omissions.
References
Al-Faris, A.R.F., 2002. The demand for electricity in the GCC
countries. Energy Policy 30
(2), 117–124.
Alrashed, Farajallah, Asif, Muhammad Zahid, 2014. Trends in
Residential Energy
Consumption in Saudi Arabia with Particular Reference to the
Eastern Province.
Al-Sahlawi, M.A., 1999. Electricity planning with demand
estimation and forecasting in
Saudi Arabia. Energy Stud. Rev. 9, 82–88.
Alyamani, Ryan, Abdulelah, Darandary, Mikayilov Jeyhun, I.,
Hasanov Fakhri, J., 2019.
Residential Electricity Price Reforms: Are Different Income
Groups and Regions
Impacted Equally? KAPSARC Publications. (Accessed on
December 18, 2019).
https://www.kapsarc.org/research/publications/residential -
37. electricity-price-
reforms-are-different-income-groups-and-regions-impacted-
equally/.
Atalla, T., Hunt, L., 2016. Modelling residential electricity
demand in the GCC countries.
Energy Econ. 59, 149–158.
Beenstock, M., Dalziel, A., 1986. The demand for energy in the
UK: a general equilibrium
analysis. Energy Econ. 8 (2), 90–98.
Campos, Julia, Ericsson, Neil R., Hendry, David F., 2005.
General-to-specific modeling: an
overview and selected bibliography. FRB Int. Finance Discuss.
Paper 838.
Chang, H.S., 1977. Functional forms and the demand for meat
in the United States. Rev.
Econ. Stat. 59, 355–359.
Chang, H.S., 1980. Functional forms and the demand for meat
in the United States: a
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772
6
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0005
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0005
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0010
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0010
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0015
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0015
https://www.kapsarc.org/research/publications/residential -
electricity-price-reforms-are-different-income-groups-and-
39. Shephard, N. (Eds.), The
Methodology and Practice of Econometrics: A Festschrift in
Honour of David F.
Hendry. Oxford University Press, Oxford, pp. 88–121.
Doornik, J.A., Hendry, D.F., 2009. Modelling Dynamic
Systems: PcGive 13. Timberlake
Consultants, LTD. London.
Doornik, J.A., Hendry, D.F., 2018. Empirical Econometric
Modelling, PcGive 15.
Published by Timberlake Consultants Ltd Timberlake Analytics,
Inc. The Loft, 2C
Blake Mews Kew Gardens Richmond, TW9 3GA, UK.
Eltony, M.N., Mohammad, Y.H., 1993. The structure of demand
for electricity in the Gulf
Cooperation Council Countries. J. Energy Dev. 18 (2), 213–221.
Engle, R.F., Granger, C.W.J., 1987. Co-integration and error
correction: representation,
estimation and testing. Econometrica 55, 251–276.
General Authority for Statistics (GaStat), 2018. Household
Income and Expenditure
Survey, 2018. (accessed on December 18, 2019).
https://www.stats.gov.sa/en/37.
Hagihara, Jun, 2013. Saudi Arabia’s domestic energy situation
and policy: focusing on its
power sector. Kyoto Bull. Islamic Area Stud. 6, 107–135.
Hasanov, F.J., 2019. Theoretical framework for industrial
electricity consumption re-
visited: empirical analysis and projections for Saudi Arabia.
KAPSARC Discussion
40. Paper. KAPSARC, Riyadh, Saudi Arabia.
https://doi.org/10.30573/KS–2019-DP66.
Hasanov, F.J., Bulut, C., Suleymanov, E., 2017. Review of
energy-growth nexus: a panel
analysis for ten Eurasian oil exporting countries. Renew.
Sustain. Energy Rev. 73,
369–386.
Hendry, D.F., Johansen, S., Santos, C., 2008. Automatic
selection of indicators in a fully
saturated regression. Comput. Stat. 33, 317–335 Erratum, 337-
339.
Karanfil, F., Li, Y., 2015. Electricity consumption and
economic growth: exploring panel-
specific differences. Energy Policy 82, 264–277.
Liddle, B., Lung, S., 2010. Age structure, urbanization, and
climate change in developed
countries: revisiting STIRPAT for disaggregated population and
consumption-related
environmental impacts. Popul. Environ. 31 (5), 317–343.
Matar, W., 2017. A look at the response of households to time-
of-use electricity pricing in
Saudi Arabia and its impact on the wider economy. Energy
Strategy Rev. 16, 13–23.
Matar, W., Anwer, W., 2017. Jointly reforming the prices of
industrial fuels and re-
sidential electricity in Saudi Arabia. Energy Policy 109, 747–
756.
Mohammadi, H., Amin, M.D., 2015. Long-run relation and
short-run dynamics in energy
41. consumption–output relationship: international evidence from
country panels with
different growth rates. Energy Econ. 52, 118–126.
Narayan, P.K., Smyth, R., 2009. Multivariate Granger causality
between electricity con-
sumption, exports and GDP: evidence from a panel of Middle
Eastern countries.
Energy Policy 37 (1), 229–236.
NIDLP, 2019. National Industrial Development and Logistics
Program. Accessed on
September 11. https://vision2030.gov.sa/en/programs/NIDLP.
Park, Joon Y., 1990. Testing for unit roots and cointegration by
variable addition. Adv.
Econometrics 8, 107–133 ed. by G. F. Rhodes and T. B. Fomby.
Greenwich, CT: JAI
Press.
Park, J.Y., 1992. Canonical cointegrating regressions.
Econometrica 60, 119–143.
Phillips, P.B., Hansen, B.E., 1990. Statistical inference in
instrumental variables regres-
sion with I(1) processes. Rev. Econ. Stud. 57, 99–125.
Saikkonen, Pentti., 1992. Estimation and testing of cointegrated
systems by an auto-
regressive approximation. Econ. Theory 8 (1), 1–27.
Salahuddin, M., Gow, J., Ozturk, I., 2015. Is the long-run
relationship between economic
growth, electricity consumption, carbon dioxide emissions and
financial development
in Gulf Cooperation Council Countries robust? Renew. Sustain.
42. Energy Rev. 51,
317–326.
Stock, J.H., Watson, M.W., 1993. A simple estimator of
cointegrating vectors in higher
order integrated systems. Econometrica 61, 783–820.
Jeyhun I. Mikayilov Jeyhun is a research fellow at
KAPSARC. His primary research interests include but are
not limited to applied time series econometrics, the eco-
nomics of energy and environment, and sustainable devel -
opment.
Fakhri J. Hasanov Fakhri is a research fellow leading the
KAPSARC Global Energy Macroeconometric Model
(KGEMM) project. His research interests and experience
span econometric modeling and forecasting, building and
applying macroeconometric models for policy purposes,
energy economics with a particular focus on natural re-
source-rich countries.
Waheed Olagunju Waheed is Director of Economic
Analysis in the Policy and Strategic Planning department at
Ministry of Energy in Kingdom of Saudi Arabia.
Mohammad H. Al-Shehri Mohammad is head of the
Modeling Department at Saudi Electricity Company.
Mohammad holds MS degree in Engineering.
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772
7
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0045
http://refhub.elsevier.com/S1040-6190(20)30064-6/sbref0050