The Electricity Journal 33 (2020) 106772

Contents lists available at ScienceDirect

The Electricity Journal

journal homepage: www.elsevier.com/locate/tej

Electricity demand modeling in Saudi Arabia: Do regional differences T

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 LE I N FO A B S T R A C T

Keywords: The paper examines the electricity demand behavior for Saudi Arabia, using annual data for the period of
Electricity demand 1990–2016, at regional level. The study finds that income, price and population are the main drivers of elec-
Price elasticity tricity demand at regional level. Although, the impacts vary across regions (central, eastern, southern and
Cointegration western), the estimated elasticities all are statistically significant, in both long and short run, and have the
Regions
expected signs for all the regions. The income, price and population elasticities range from 0.10 to 0.93, from
Saudi Arabia
-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
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
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).

⁎
Corresponding author.
E-mail address: jeyhun.mikayilov@kapsarc.org (J.I. Mikayilov).

https://doi.org/10.1016/j.tej.2020.106772

1040-6190/ © 2020 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/BY/4.0/).
J.I. Mikayilov, et al.
A regional approach for KSA constitutes a novel and significant
contribution to the literature on electricity demand in the following
ways. First, to the best of our knowledge, there are very few to none
studies that examine the determinants of electricity demand at the re-
gional level for KSA. Alyamani et al. (2019) considered regional aspects
of the electricity consumption. However, they only considered re-
sidential electricity consumption and did just a descriptive analysis.
Diabi (1998) examined total electricity consumption for the five 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-
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
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
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.
2
The Electricity Journal 33 (2020) 106772
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,
Alyamani et al. (2019) is descriptive-type study and does not estimate
elasticities. Diabi (1998) studied total electricity consumption 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
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; 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
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.
J.I. Mikayilov, et al.
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)
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, 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
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
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 :
3
The Electricity Journal 33 (2020) 106772
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 0 + α1 Income + α2 Population + α3 Price + e
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 α1, and α2 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 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
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 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

Fig. 1. Shares of regional electricity demands in total, %.

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
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).
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 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. Empirical estimation results 6.3. Long-run estimation results

6.1. Unit root test results After concluding the cointegration relationship among the variables
the long-run estimation results can be interpreted so they are not
As is standard practice in the literature, the main concern lies in spurious. The long-run estimation results are provided in Table 3. To
dealing with stochastic properties of the variables and hence the have an idea of the impacts in terms of magnitude ranges, we reported
Augmented Dickey-Fuller (ADF) Unit Root test was employed for each the 99 % confidence intervals for the estimated coefficients/elasticities.
variable in levels and first differences. In unit-root testing, the max- The interval estimator provides more information about the re-
imum lag length is taken to be two and optimal lag is chosen based on presentativeness quality of the estimated coefficient, in addition to
Schwarz information criterion (SIC). Table 1 displays the results of the delivering the limits of the impact.
Unit Root tests. Based on the estimation results presented in Table 3, one can see
Briefly, based on the results displayed in Table 1 we can conclude that for all the regions, the impacts of income, price and population are
that all variables are integrated of the first order [I(1)] and there is economically meaningful as they take right signs and statistically sig-
potential for cointegration relationship. Hence, we can proceed to the nificant at 1% level. As can be seen from Table 3, the income elasticity
of electricity demand varies across regions, demonstrating more stable
Table 1 behavior in the Western region and wider ranged in the Central region.
ADF Unit Root Test results. The highest income elasticity found in Central region, while it has the
smallest value in the Eastern region. Overall, the income elasticity
Region Variable (in logs) Level First Difference
ranges from 0.102 to 0.931 across the regions.
National gdp −0.948 −4.269*** Price elasticity of electricity demand is found to be ranging from
di −0.138 −5.147*** -0.607 to -0.060 across the regions.
Central dele_coa −0.976 −4.962***
The impact of population demonstrates more similar path across
pele_coa −1.274 −4.514***
pop_coa −0.271 −5.618*** regions, having differences as well. Numerically the population elasti-
Eastern dele_eoa −1.576 −4.403*** city ranges from 0.243 to 0.947.
pele_eoa −1.782 −4.465***
pop_eoa −0.582 −3.806*** 6.4. Short-run estimation results
Western dele_woa −0.903 −6.994***
pele_woa −1.898 −4.438***
pop_woa −0.139 −4.668*** For the short-run analysis, we applied Gets to ECM to determine the
Southern dele_soa −0.444 −5.110*** relationship for each region. We start with a general ECM specification,
pele_soa −1.327 −4.534*** which includes the error correction term (ECT), contemporaneous va-
pop_soa −0.670 −4.476***
lues of all independent variables and 2 lags of all the variables. We then
Note: dele = demand for electricity, COA = central operating region, exclude variables from the analysis based on the test proposed by the
EOA = eastern operating region, SOA = southern operating region, methodology and end up with the final short-run specification (e.g., see
WOA = western operating region, pop = population, pele = electricity price; Campos et al., 2005 inter alia). We are not reporting the General Un-
*** =Significant at the 1% level, intercept only case is chosen based on the restricted Models here to save space, but they are available from the
analyses. authors upon request. Table 4 documents the final ECM specification

4
J.I. Mikayilov, et al. The Electricity Journal 33 (2020) 106772

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 As implied by Table 3, the long-run income elasticities range from

Long-run Estimation results. 0.10 to 0.93 across the regions, the long-run price elasticities range
Regressor Region
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
COA EOA SOA WOA elasticities for income and population, and negative elasticities for
prices that we discussed in the Theoretical Framework section. In terms
Income* (0.403, 0.931) (0.102, 0.204) (0.120, 0.360) (0.426, 0.470)
Price (−0.580, (−0.628, (−0.132, (−0.607,
of magnitudes, our long-run income elasticities are smaller than finding
−0.362) −0.356) −0.060) −0.427) of Al-Faris (2002), which is the only study devoted to total electricity
Population (0.243, 0.661) (0.771, 0.947) (0.335, 0.645) (0.776, 0.840) 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
Notes: Dependent variable is dele_i and i takes COA, EOA, SOA and WOA, re- than his finding. When it comes to long-run price elasticity, the found
spectively; * for the Central and the Southern regions GDP used as an income
range is in line with the expectation for the developing country case.
measure; For the Eastern and Western regions Disposable Income is used as
For example, Atalla and Hunt (2016) estimated the long-run price
income proxy; Numbers in parentheses are 99 % confidence intervals for the
obtained elasticities. elasticity for residential demand to be -0.16, which is well contained in
our interval. In addition, since the previous studies investigated the
and the post-estimation test results. electricity demand modeling at overall country level, our results are not
As the table demonstrates the short-run final specifications pass all directly comparable with their results. The country level elasticities can
the diagnostic and misspecification tests, hence the results are inter- be seen as the average representative for the whole country, while the
pretable. Two observations are worth stating. First, all the remained regional ones might take into account the region-specific features.
regressors in the final ECM have economically meaningful and statis- With an income elasticity of 0.93 (the upper bound of the con-
tically significant impact on the electricity consumption in all regions. fidence interval), electricity demand in the Central region is the most
Second, this is true for the contemporaneous values of all the drivers, sensitive to changes in income. The Eastern region is the most sensitive
i.e., the income, price and population. to changes in population as the population elasticity of electricity de-
The estimated Speed of Adjustment (SOA) coefficients (coefficient mand is the highest at 0.95 (the upper bound of the confidence in-
of the ECT term) are statistically significant and negative in all the terval).
regions, confirming the existence of stable long-run relationship among With a price elasticity of −0.61 (The lower bound of the confidence
the variables. 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
7. Discussion of the results of all the elasticities for each region, income has the largest impact on
electricity demand in the Central region, population has the largest
In this section, we present the key results and insights from our impact in the Eastern region, while price has highest impact in the
analysis on regional electricity demand. The key results are summarized Western region.
in Tables 3 and 4. Table 3 displays the estimated long run elasticities for Comparing the elasticities across regions, we observe that in abso-
each region and Table 4 displays the short run elasticities. lute terms, the price elasticity is lowest in the Southern region (-0.13,

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) *
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.

5
J.I. Mikayilov, et al.
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
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 reinforced 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-
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
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 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
6
The Electricity Journal 33 (2020) 106772
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
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
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-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

reply. Rev. Econ. Stat. 62, 148–150. order integrated systems. Econometrica 61, 783–820.
Chang, H., Hsing, Y., 1991. The demand for residential electricity: new evidence on time-
varying elasticities. Appl. Econ. 23, 1251–1256.
Chang, Y., Choi, Y., Kim, C.S., Miller, J.I., Park, J.Y., 2016. Disentangling temporal pat- Jeyhun I. Mikayilov Jeyhun is a research fellow at
terns in elasticities: a functional coefficient panel analysis of electricity demand. KAPSARC. His primary research interests include but are
Energy Econ. 60, 232–243. not limited to applied time series econometrics, the eco-
Diabi, A., 1998. The demand for electric energy in Saudi Arabia: an empirical in- nomics of energy and environment, and sustainable devel-
vestigation. OPEC Rev. 22 (1), 13–29. opment.
Dickey, D., Fuller, W., 1981. Likelihood ratio statistics for autoregressive time series with
a unit root. Econometrica 49, 1057–1072.
Doornik, J.A., 2009. Autometrics. Chapter 4 in In: Castle, J.L., 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 Fakhri J. Hasanov Fakhri is a research fellow leading the
Cooperation Council Countries. J. Energy Dev. 18 (2), 213–221. KAPSARC Global Energy Macroeconometric Model
Engle, R.F., Granger, C.W.J., 1987. Co-integration and error correction: representation, (KGEMM) project. His research interests and experience
estimation and testing. Econometrica 55, 251–276. span econometric modeling and forecasting, building and
General Authority for Statistics (GaStat), 2018. Household Income and Expenditure applying macroeconometric models for policy purposes,
Survey, 2018. (accessed on December 18, 2019). https://www.stats.gov.sa/en/37. energy economics with a particular focus on natural re-
Hagihara, Jun, 2013. Saudi Arabia’s domestic energy situation and policy: focusing on its source-rich countries.
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
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. Waheed Olagunju Waheed is Director of Economic
Karanfil, F., Li, Y., 2015. Electricity consumption and economic growth: exploring panel- Analysis in the Policy and Strategic Planning department at
specific differences. Energy Policy 82, 264–277. Ministry of Energy in Kingdom of Saudi Arabia.
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
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.
Mohammad H. Al-Shehri Mohammad is head of the
Energy Policy 37 (1), 229–236.
Modeling Department at Saudi Electricity Company.
NIDLP, 2019. National Industrial Development and Logistics Program. Accessed on
Mohammad holds MS degree in Engineering.
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. Energy Rev. 51,
317–326.
Stock, J.H., Watson, M.W., 1993. A simple estimator of cointegrating vectors in higher