Proc.

of the 4rth International Conference on Power Generation Systems and Renewable Energy Technologies (PGSRET)
10-12 September 2018, Islamabad, Pakistan

An Elasticity based Deterministic Study of

Relationship between Factors Effecting
Domestic Energy usage in Pakistan
Hafiz Muhammad Bilal Ahmed Abdul Kashif Janjua Dr. Muhammad Bilal Khan
United states Pakistan Center for United states Pakistan Center for
Advanced Studies in Energy Advanced Studies in Energy United states Pakistan Center for
Advanced Studies in Energy
NUST NUST
Islamabad, Pakistan Islamabad, Pakistan NUST
[email protected] [email protected] Islamabad, Pakistan
[email protected]

Abstract— Pakistan’s energy sector is in chaos, policy long term policy and wrong investments on rental type of
to increase dependency on fossil fuel has played a big electricity the power sector is facing the crisis it never
part in increase of circular debts. Policy making depends on
saw before in the country. Six to eight hours of
various factors and forecasting correct requirements for
energy is one of the key factors for policy making thus electricity outages in winters and the situation is worse
Predicting energy demand and considering the impact of in summers for urban area while rural areas suffer more.
macroeconomic indicators on household power consumption Similar is the case with Nigeria and Egypt,
was always part of effective energy policy planning. The Consequently the social standards of the country are
paper presents a suitable model that contains macro-
declining day by day and the suffering of economic
indicators like GDP, real price, consumers, population as
independent variable and sales of electricity unit as growths is eminent, moreover the industrialization is also
dependent variables, through the use of the method strained with similar issues similar issues which can be
revealing the econometric relationships. Particularly observed in all the aforementioned countries [3].
considering the effects of elasticities that are influencing
the model for the prediction of household electricity demand The Energy policy decisions have a profound and long-
are modeled. The study period for this research was term impact on the structure of the dominant energy
1980-2015. The paper concludes the factors which are systems. Thus knowledge of energy market mechanisms
responsible for increase in energy demand of the country
which may help government planning commission to is vital. Elasticity is one of the key factors to observe
improve the energy policies for future. such market behaviors, the elasticity estimates provide
information on the sensitivity of consumer behavior to
Keywords— Regression model, cointegration, economic the evolution of the main explanatory variables Energy
growth, Developing Countries, Pakistan prices, income, energy use of the capital stock. During
the energy policy formation, it is a great challenge for
I. INTRODUCTION energy economists to find out demand elasticities that
All the energy mix of the world is in a transition period are great indicators for long-lasting policy design. [4].
with the fluctuating prices of oil introduction of new
Energy consumption in Pakistan increased dramatically.
protocols every day for green energy production in the
In previous decade due to population growth and the
world is now a very common phenomenon [1]. While
economy, which has led to an energy deficit, and
the world is busy making its energy green there still are
confusion can still be felt today. Reliability of energy
economies and power systems who even fail to meet the
demand elasticities is not always useful because it
most basic need of the consumer i.e. constant supply of
takes a long time to get an optimized model [5].
electricity. Getting a very clean electricity without any
Quantitative analysis of energy demand seems to
harmonics is a myth, in some developing countries such
dominate applied research. Recently, new techniques
as Nigeria, Egypt, Pakistan, India and some other
have become increasingly popular and used by the
developing countries in past even china had a mass area
researcher, while at the same time trend of demand
which no electricity was available or electricity outages
elasticities estimation seems to have declined. But both
were a constant occurrences until government invested in
decision makers and energy modelers who offer model-
DG sources and d big dams for water and energy needs.
based ideas depend on reliable estimates of the
Other countries despite knowing the consequences were
sensitivity of energy demand in response to changes in
unable to implement such policies. For example in
key variables that make up energy demand[6]–[8].
Pakistan due to many reasons significantly due to lack of

978-1-5386-7027-9/18/$31.00 ©2018 IEEE

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 29,2020 at 10:05:35 UTC from IEEE Xplore. Restrictions apply.
Objectives of the proposed research is to apply and demand elasticities is not always useful because it
demonstrate the techniques to estimate the elasticities of takes a long time to get an optimized model [5].
domestic sector. One of the main concerns of Quantitative analysis of energy demand seems to
Pakistan's electricity generation is the high rate of dominate applied research. Recently, new techniques
domestic demand growth and the inefficient use of have become increasingly popular and used by the
electricity in other sectors researcher, while at the same time trend of demand
energy economists to find out demand elasticities that elasticities estimation seems to have declined. But both
are great indicators for long-lasting policy design. [4]. decision makers and energy modelers who offer model-
based ideas depend on reliable estimates of the
Energy consumption in Pakistan increased dramatically. sensitivity of energy demand in response to changes in
In previous decade due to population growth and the key variables that make up energy demand[6]–[8].
economy, which has led to an energy deficit, and
confusion can still be felt today. Reliability of energy
along with load shedding. Primarily, model defines the
II. METHDOLOGY linear relationship between an explanatory variable and
To develop dynamic model, that can establish long run co- dependent variable [10].
integration among potential variables in which domestic Further analyzing the data presented in Fig.1. We can see
electricity demand is dependent variable. Factors that are overall growth potential in the system likewise in all sectors
responsible for electricity demand in domestic sector are the for a better understanding curves were fit using regression
key components to investigate. Regression analysis of analysis on the shown data to observe the trends for future.
Factors involving long-run relationship in domestic sector In the following figure, it is time in year starting from year
electricity demand can be shown by general equation. 1970 and age is the agriculture GWh requirements of
(1) electricity.
Where S shows electricity sales in domestic sector. In this The function fit for the regression analysis is a polynomial
regression equation S is dependent variable. βi represents of degree 3 with an R-sqr value of 0.917 has the best fit on
respective explanatory variable coefficient derived from the data used for testing and validation having 3 subsequent
regression analysis. Xi represents independent variables data. The function and the coefficients are as follows:
included in this model. Number of domestic customers, Real p1*x^3 + p2*x^2 + p3*x + p4
price of KWh in domestic customers, income, inflation and p1 = 0.06292 (-0.104, 0.2299)
Gross domestic product (GDP) are the independent p2 = -1.438 (-13.37, 10.49)
variables included in this model to check the sensitivity of p3 = 187 (-55.65, 429.7)
these variables with dependent variables. ε is error term p4 = 361.5 (-969.9, 1693)
summing the combined effect of variables that are not With 95% confidence bounds. Similarly the summation of
included in the model but can still influence the dependent all the different sectors are also regressed under similar
variable, and c is constant. Dummy variables also used in conditions results are being shown in the figure
that indicate either of the followings i). The government
policy (closed or open) ii). Presence or absence of quality in
service. Its value is either 0 or 1.

Fig. 3 Best Fitted Trend

The cumulative GWh power shows a perfect increasing
trend over the years and exponential trends are the better
representation of such trends fit similarly
f(x) = a*exp(b*x)
Where
a = 6289 (5587, 6991) b = 0.06271 (0.05988, 0.06555)
Using regression analysis we can get a curve for next 50
years which is as follows the results are very much
coinciding with the current statistics of the GWh production
in the country [11].
Fig. 1 Electricity Demand History in four Economic groups
In Pakistan, energy consumption in different sectors can be
shown in figure 1. Eviews and matlab are the major
software used to analyze and develop a model in this
research. Pakistan’s Economic Survey and Power statics are
the main sources to collect data used in this model [9]. To
get a better insight of data, it was modified by summing up
the domestic energy consumption in different Power entities

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 29,2020 at 10:05:35 UTC from IEEE Xplore. Restrictions apply.
 method can be used to determine the long-run association
between the variables.
Table 1 Maximum likelihood estimation

Fig. 4 Growth representation over the Years

In this model null hypothesis of (r=0) at 5% level rejected as
determined by trace test. And a number of cointegration
equations ranked 4 as confirmed by both trace and
maximum eigenvalue as shown in Table 1. These tests
showed no deviation from theoretical model assumption.
Once the cointegration among the endogenous variables
established, we can develop an error correction model. In
error correction model, we ranked our model at 4
cointegration equations level. This method suggests
different models with a coefficient and standard error [13].
But we have chosen the only model with the demand for
electricity as the dependent variable. To remove
insignificant values we have to know p-value of these
coefficients. At this stage, all values are no longer at level
but rather with a difference to estimate error correction.
Restricted VECM used to quantify short-term elasticities.
Lag length used started by Lag 5 and after selecting the first
model with dependent variable i.e. sales of energy. The
Fig. 5 Anatomy of Econometric Model insignificant coefficient can be removed by p values. To
make the equation balance, we added some variables
without lag. To check the Stability parameter, we applied
III. RESULTS AND DISCUSSION CUSUM test which shows no sign of instability in
All data used in this model has a unit root at level. These parameters.
non-stationary values converted to stationary values after
first differencing. Without converting to stationary values
regression results can be spurious that seems plausible but
of no use. Non- stationary behavior was checked on each
individual time series. For this an Augmented Dickey–Fuller
test (ADF) test was applied for each non-breaking time
series. After first differencing null hypothesis rejected at 5%
level.
All data used in this model has a unit root at level. These
non-stationary values converted to stationary values after
first differencing. Without converting to stationary values
regression results can be spurious that seems plausible but
of no use. Non- stationary behavior was checked on each
individual time series. For this an Augmented Dickey–Fuller
test (ADF) test was applied for each non-breaking time Fig. 6 CUSUM test for estimated ECM
series. After first differencing null hypothesis rejected at 5% Different tests were conducted to check stability in the
level. Cointegrating vectors can be find out if all model. To check normal distribution of residuals in
endogenous variables are of the same order. For this endogenous variables Jarque-bera statistics and
research, we used maximum likelihood estimation corresponding probability measured found to be 5.2198
elaborated by Johansen and Juselius (1988) [12]. This (0.07353), as its probability is >5% which means we can
accept null hypothesis. In this test null hypothesis says that

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 29,2020 at 10:05:35 UTC from IEEE Xplore. Restrictions apply.
residual are equally distributed. Heteroskedasticity test done
on model time series to check there is no ARCH effect. As p REFERENCES
found to be >5% level, we accepted null hypothesis (Ho)
[1] H. M. Bilal and A. Z. Khan, “Economic planning of network for
[14]. Null hypothesis says there is no arch effect when p> integration of renewable: A review,” 2015 Power Gener. Syst. Renew.
5%. Breusch-Godfrey Serial Correlation LM test was done Energy Technol., pp. 1–3, 2015.
to check that there is no serial correlation which is null
hypothesis. We accepted null hypothesis as values found to [2] A. K. Janjua and S. T. Jan, “Smart Nano Grids a novel approach For
Power Sharing among Pakistan ’ s Domestic Consumers A case study for
be greater than 5%. Summary of tests can be seen in table 2.
domestic local power profiling for smart Nano grids.”
This theoretical economic model passed all the diagnostic
tests conducted during developmentEase of Use [3] S. Farooq, “Smart Grid — Energizing Towards a Greener Future,” 2015.
Table 2 Diagnostic Tests
[4] B. Cleary, a Duffy, a O’Connor, M. Conlon, and V. Fthenakis,
“Assessing the Economic Benefits of Compressed Air Energy Storage for
Mitigating Wind Curtailment,” Sustain. Energy, IEEE Trans., vol. 6, no. 3,
pp. 1021–1028, 2015.

[5] I. A. Ezenugu, S. C. Nwokonko, and I. Markson, “Modelling and

Forecasting of Residential Electricity Consumption in Nigeria Using
Multiple and Quadratic Regression Models,” Am. J. Softw. Eng. Appl., vol.
Behavior of variables in future can be predicted by impulse 6, no. 3, pp. 99–104, 2017.
response function by one standard deviation positive shock. [6] D. Zhang, M. Xiao, X. Yang, and Y. He, “The analysis of
In figure 4, one standard deviation positive shock of the manufacturing PMI potential trends of the US, EU, Japan and China,”
population was given to other variables. As can be seen in Procedia Comput. Sci., vol. 55, no. Itqm, pp. 43–51, 2015.
this Figure 4 Price of KWh showing a negative response. It
means domestic consumers are not affected by price per unit [7] N. H. Mirjat, M. A. Uqaili, K. Harijan, G. Das Valasai, F. Shaikh, and
M. Waris, “A review of energy and power planning and policies of
in the longer term. GDP and inflation have long-term effect Pakistan,” Renew. Sustain. Energy Rev., vol. 79, no. May, pp. 110–127,
on energy demand. Income of household has a short term 2017.
positive effect on demand [15].
[8] R. L. Hetzel, “Estimating Money Demand Functions,” J. Money, Credit
Bank., vol. 16, no. 2, p. 185, 1984.
IV.CONCLUSION
[9] S. H. E. Saleh, A. N. Mansur, N. A. Ali, M. Nizam, and M. Anwar,
Results of this model show the relation between electricity “Forecasting Of the Electricity Demand in Libya Using Time Series
demand and econometric variables. Some variables Stochastic Method for Long- Term From 2011-2022 .,” vol. 3, no. 5, pp.
show less significance as compared to others. Price of per 12529–12536, 2014.
kWh in domestic sectors has a short-term effect. For [10] M. Praptiningsih and Y. Tanoto, “Forecasting Residential Electricity
impulse response function (IRF) in Fig. 7, results are based Consumption in Indonesia,” Proc. Yogyakarta Econ. Bus. Int. Conf. 2013,
on long- term behavior rather transitory, the price is pp. 35–50, 2013.
showing a negative response, which means price is not
[11] E. Ortjohann, W. Sinsukthavorn, N. Hamsic, A. Schmelter, A. Mohd,
showing a long- term relationship with demand. Demand D. Morton, “An Innovative Simulation Approach Using Dynamic-RMS
shows relatively more sensitivity to GDP and power model for Hybrid Isolated Mini-Solars”, The Great Wall World Renewable
network expansion to distant areas in Pakistan and have a Energy Forum and Exhibition 2006, Beijing, China, (2006)
long-term significant coefficient. The government is also [12] E. Ortjohann, P. Wirasanti, M. Lingemann, W. Sinsukthavorn,
encouraged to foresee the risk that might arise in relation S.Jaloudi, D. Morton, “Multi Level Hierarchical Control Strategy for Smart
to energy policy. This model proposes utilization of these Solar Using Clustering Concept.”, ICCEP- International Conference on.
variables into consideration and encourages the CLEAN ELECTRICAL POWER Renewable Energy Resources Impact,
Italy, June (2011)
government to invest in long-term energy plans.
[13] M. Lingemann, E. Ortjohann, W. Sinsukthavorn, S. Jaloudi, D.
Morton, “Clustered Multi-level Hierarchy for Secondary Power System
Control”, IASTED- Innovative Smart Solar Technologies for Sustainable
Power and Energy Systems, Thailand, November (2010)

[14] E. Ortjohann, W. Sinsukthavorn, A. Mohd, N. Hamsic, M. Lingemann,

D. Morton, “A Dynamic-RMS Modeling Method for Distributed
Generation”, IEEE Power Engineering Society 2009 General Meeting,
Calgary, Alberta, Canada, (2009)

[15] E. Ortjohann, W. Sinsukthavorn, A. Mohd, N. Hamsic, A. Schmelter,

D.Morton, “Dynamic-RMS Modeling of Distributed Electrical Power
Supply Systems”, IEEE PES Transmission and Distribution, Chicago,USA,
(2008)

Fig. 6 Impulse responses of electricity

consumption

Authorized licensed use limited to: INTERNATIONAL ISLAMIC UNIVERSITY. Downloaded on July 29,2020 at 10:05:35 UTC from IEEE Xplore. Restrictions apply.