An empirical study of electricity theft from electricity distribution companies in Pakistan.
Jamil, Faisal ; Ahmad, Eatzaz
Electricity theft is a common problem in many countries and energy
worth billions of dollars is stolen annually from electricity grids. The
problem has socioeconomic, political, environmental and technical roots,
but the solution is generally sought solely through technical measures.
This paper empirically investigates the effects of various factors
including electricity price, per capita income, probability of
detection, fines collected from offenders, weighted temperature index
and load shedding, that may explain the theft. The study employed annual
panel data obtained from nine electricity distribution companies in
Pakistan for the period 1988-2010. The study estimates the Fixed Effects
models through the least squares dummy variable (LSDV) technique and
Generalised Method of Moments (GMM). Our results indicate that per
capita income has significant negative and electricity price a positive
effect on electricity theft with sufficiently high coefficient values.
The probability of detection variable appears with a positive sign in
both estimations indicating a poor deterrence. The results of LSDV show
a positive impact of fine on conviction on electricity theft. But in GMM
estimation, this variable appears with a right sign. The results from
both models are robust in the case of load shedding and temperature
variables. The findings show that economic variables are most
significant in explaining electricity theft. The findings may also be
applicable in other developing countries where hefty amounts of revenues
are lost due to electricity theft.
Keywords: Electricity Theft, Fixed Effects Model, Pakistan
1. INTRODUCTION
Electricity theft is common in many countries and a considerable
amount is stolen every year from electricity grids. It deteriorates the
financial condition of the utilities, curtails new investments for
capacity development of electricity industry that eventually leads to
electricity shortage [Jamil (2013)]. If the electric utilities concerned
are public monopolies, they may seek public investment and resort to
government subsidies for their financial survival and for continued
supply of electricity maintaining the status quo. Financial condition of
few electricity distribution companies in Pakistan is extremely poor as
the revenues from sale of electricity fall short of the supply cost
[Kessides (2013)]. Huge distribution losses adversely affect the
utilities' profitability and consequently the quality of service.
These losses include technical and non-technical losses where
non-technical losses mainly constitute electricity pilferage and theft.
The financial loss due to electricity theft alone accounts for hundreds
of millions of dollars annually [see, for example, Smith (2004); Lovei
and McKechnie (2000)]. The overall mismanagement of power sector
including the heavy losses and theft inter alia resulted in accumulated
circular debt of over Rs 850 billion in 2012 [IPP (2009); FODP (2010),
Planning Commission (2013)].
Pakistan is facing acute electricity shortage and the honest
consumers have to pay heavily for quite irregular supplies. The
electricity tariff rates for consumers are essentially set on the higher
side due to widespread electricity theft. Therefore, it is pertinent to
put efforts to rectify this menace for the electricity sector.
Electricity theft has socioeconomic, political and technical basis, but
the solution is generally sought solely through technical measures. In a
recent study on electricity theft in agricultural sector in Rajasthan,
Katiyar (2005) finds that electricity theft is not possible to be
controlled in agriculture sector through a purely technical approach.
The role of socioeconomic and institutional factors is typically
under-rated in explaining and handling electricity theft issue. There
are a few contemporary studies that discuss theft and corruption in
electric utilities [for example, Clarke and Xu (2004); Smith (2004);
Estache, et al. (2006); Bo and Rossi (2007); Gulati and Rao (2007);
Nakano and Managi (2008) and Nagayama (2010)].
There is vast literature on economics of crimes and overall
corruption, however, few studies examine corruption particularly in
energy sector [for example, Clarke and Xu (2004); Bo and Rossi (2007)].
Using enterprise level data on bribes paid to electric utilities in 21
transition economies from Eastern Europe and Central Asia, Clarke and Xu
(2004) explore how characteristics of utilities taking bribes and the
firms paying bribes affect corruption in the sector. The study favours
privatisation as bribe is found more prevalent in public owned
utilities; bribe is positively related with capacity constraints and
negatively related with level of competition. Bo and Rossi (2007) trace
link between inefficiency and corruption by using a dataset comprising
firm-level information on 80 electricity distribution firms in Latin
America for the period 1994-2001. The study finds that corruption makes
the firms inefficient, as such firms employ relatively more inputs to
produce a given level of output.
Smith (2004) examines electricity theft determinants, its
consequences, and suggests some remedial measures. The study shows that
electricity theft is strongly related to governance indicators, and that
higher levels of electricity theft persist in countries with less
effective accountability, political instability, low government
effectiveness and higher corruption. He suggests that electricity theft
can be reduced primarily by applying a mix of technical solutions such
as tamper-proof meters associated with managerial methods such as
inspection and monitoring, and overall restructuring the electricity
sectoral ownership and regulation. In another recent study, Nagayama
(2010) identifies the effects of power-sector reforms on the sectoral
performance indicators (for instance, installed capacity, transmission
and distribution losses) and finds that reform variables such as the
entry of Independent Power Producers (IPPs), unbundling of generation
and transmission, establishment of regulatory agencies, and the
introduction of a wholesale spot market lead to the increased generation
capacity as well as reduced transmission and distribution loss in the
respective regions. On the whole, literature focuses mainly on supply
aspects of electricity theft and identified that poor governance, lack
of competition and inefficiency are major causes of electricity theft.
This study is based on the argument that electricity theft is a
multidimensional issue and ought to be investigated from a broader
perspective. We examine the role of various factors that affect
electricity theft by using panel data of electricity distribution
companies in Pakistan for the period 1988-2010. Each of the distribution
company serves its customers in a specific region of Pakistan. The data
shows that there are startling differences of electricity pilferage
rates in different companies/regions. We explore the determinants of
electricity theft in order to explore answers to a number of questions
such as the following.
* Is electricity theft affected by the economic activity?
* How responsive are the consumers to the electricity tariff that
is, if tariff rate increases, the consumers reduce their electricity
consumption or opt for electricity theft? Answer to this question may
depend on price elasticity of electricity demand and consumers'
expected risk of detection. (1)
* Are the offenders responsive to the probability of detection and
magnitude of fines?
* Does the climate affect the electricity theft?
* Whether quality of electricity service affects the consumer
behaviour regarding their theft decision?
Our empirical analysis comes up with answers to these questions. We
employed Fixed Effects modelling. The Fixed Effects models are estimated
using least square dummy variables (LSDV) and generalised method of
moments (GMM) methods. Our results indicate that per capita income has
significant negative and electricity price has positive effect on
electricity theft or pilferage with high magnitudes of coefficients.
Similarly, temperature variable has significant positive impact on
electricity theft. However, the probability of detection and penalty for
the offence i.e. fine variables do not perform consistently in all the
models, partly due to poor monitoring and the law implementation and
partly due to data quality. The fine on theft detection is found
significant with negative sign.
The remainder of this paper is organised as follows. Section 2
briefly describes the electricity theft situation in Pakistan. Section 3
provides the conceptual framework and Section 4 presents the model and
variables. The econometric methodology is given in Section 5. The
results are discussed in Section 6, while Section 7 concludes the paper.
2. ELECTRICITY THEFT SITUATION IN PAKISTAN
The study investigates electricity theft and estimates the
contributions of factors by using a dataset of electricity distribution
companies operating in Pakistan. There are nine distribution companies
operating in the country including, Islamabad Electricity Supply Company
(IESCO), Lahore Electricity Supply Company (LESCO), Gujranwala Electric
Power Company (GEPCO), Faisalabad Electricity Supply Company (FESCO),
Multan Electric Power Company (MEPCO), Peshawar Electricity Supply
Company (PESCO), Quetta Electricity Supply Company (QESCO), Hyderabad
Electricity Supply Company (HESCO) and Karachi Electric Supply Company
(KESC). These distribution companies are public monopolies with the
exception of KESC, which has been privatised since 2005 and operates in
metropolitan Karachi and has exclusive rights to supply power in its
jurisdiction.
A region of operation for each distribution company is established
by the government and these regions possess different social, political
and economic characteristics. This is why the likelihood and extent of
theft, its detection and conviction rate and modes of theft differ among
the utilities. In spite of such diversity, moderate to high rate of
theft and moderate to low detection rates prevail in most of the
distribution companies. The intensity and incidence of electricity theft
may differ in different parts of the country, whereas electricity theft
is a common practice in most places. The average distribution losses in
2012-13 were found to be as low 9.5 percent in 1ESCO to be as high as 36
percent in PESCO. The transmission and distribution losses of KESC
exceed 40 percent for some of the years [KESC (2006)]. On average, 20-25
percent of total electricity generated in Pakistan is marked as
distribution losses. Power theft has been so serious issue in Pakistan
that the government had to deploy army to recover electricity charges of
distribution companies in 1999. Table 1 shows the disparity in
electricity losses among all the distribution companies.
Table 2 gives a glimpse of the theft detection, penalty and
recovery against the fines imposed. There are differences in electricity
theft, conviction rates and law enforcement among the utilities and
regions. The situation is worse in KESC, PESCO and HESCO with high
losses, high detections and low recovery of fines imposed. The situation
is better in utilities of central Punjab like IESCO, FESCO and GEPCO,
where the losses fall in the range of 10-13 percent during the period
analysed.
3. CONCEPTUAL FRAMEWORK
The economics of electricity theft is essentially concerned with
the cost and benefits of limiting the non-violent crime of electricity
theft from the electricity distribution systems. The benefits of
curtailing theft are in the form of increased revenues of utilities and
consequently, improved electricity supply for the consumers. The
potential costs include surveillance expenditures of utilities, rewards
to monitors, and price incentives to consumers. Corruption and bribe are
common in regions where electricity theft is widespread. The factors
that entrench corruption and electricity theft are their beneficial
features for consumers in terms of lowering electricity cost as well as
private illegal incomes for corruptible employees of utilities. The
ultimate victim is the utility/government and honest consumers at large.
Economic theory suggests that crime is committed only if the gain
from offence exceeds the expected cost. The economic cost-benefit
analysis of electricity theft aims to develop optimal public and private
policies to combat this crime. From enforcement point of view,
individuals can be deterred either by increasing the fine or by
increasing the probability of detection. The increase in probability of
detection and conviction is costly as it essentially requires the
utilities to increase surveillance expenditure. Alternatively, utilities
can increase the expected cost of electricity theft by increasing the
fine for convicted [see, for instance, Becker (1968); Becker and Stigler
(1974)]. The study proposes that the probability of detection and
conviction may complement the amount of fine in deterring individuals
from committing the crime. Theft comprises of the incidents where
distribution companies fail to recover their receivables due to illegal
abstraction of electricity by consumer, and improper recording and/or
reporting by their employees. As a result, the actual receivables are
not recovered. Electricity theft harms the financial condition of
electric utilities and negatively affects future investments in power
sector.
Electricity industry in most of developing countries is
characterised by extensive public interventions sometimes to pursue
their social, economic and political objectives. The result is
widespread corruption in the sector, inefficiencies at the generation
and distribution levels and poor financial performance of utilities.
Joseph (2010) argued that getting the electricity prices right may not
suffice in reducing the financial instability of utilities, when the
system is burdened with electricity theft and corruption. An equally
pertinent issue in most developing countries is non-payment of due
electricity charges by customers.
Electricity is generated at various power stations, which are
generally located at distances from the load centres or end-users. It is
then transported to end-users through wires and conductors. Electricity
delivered by utility may differ from electricity billed due to technical
and non-technical losses. When electricity passes through a wire, a
fraction is lost due to the resistance of the conductor and stepping up
and down of voltage and this is generally called technical loss.
Non-technical losses mainly constitute electricity theft. Electricity
theft can take place through a number of means and ways. Electric
utilities charge electricity on the basis of meter readings at the
consumers' interface. The distribution lines of the utilities lie
open and hence the chances exist of consumers' illegally
abstracting electric power through by-passing or even with tempering the
meter.
In order to supply electricity to its consumers, utility delegates
to employees various activities, such as repairing and maintenance,
theft identification and electricity retailing. Corruption facilitates
electricity theft wherein consumer and utility employee collude for
personal gains ultimately causing a loss to the utility and public at
large. The utility employees directly interact with the consumers and
hence may help consumers in hiding the actual electricity consumption by
receiving nominal bribes from them. Both the corrupt employees and
consumers gain through this illicit relationship.
We are primarily concerned with the cost and benefits of limiting
electricity pilferage among consumers. The benefits of curtailing theft
are increased revenues of utilities and improved investment. The
potential costs may be increased surveillance expenditures as well as
rewards and price incentives. Smith (2004) emphasised the link between
corruption and electricity theft and states that low transmission and
distribution losses (around 6 percent) are most common in countries with
low corruption perception like Belgium, Finland and Germany and while
higher losses (around 30 percent) are most common in countries with high
corruption perception like Albania, Bangladesh, Haiti, India and
Pakistan. The study further identifies that electricity theft is highly
correlated with all governance dimensions, such as civil rights,
democratic institutions arid accountability. The deterrent measures
adopted for curbing the electricity theft are mainly technical such as
introduction of advanced electricity meters. To deal with the
multidimensional inter-linked aspects, this study is structured to
specify a model of electricity theft by identifying explicitly the major
economic and institutional policy variables to combat electricity theft
in Pakistan.
4. MODEL AND VARIABLE CONSTRUCTION
This section highlights the factors that might affect electricity
theft in Pakistan. We employ the most relevant variables as regressors
comprising of utility-specific variables as well as country-specific
variables taken as common for all utilities. The analysis is based on a
dynamic panel model for electricity theft using panel data for nine
electricity distribution companies in Pakistan. The general regression
equation is as follows.
[TH.sub.i,t] = f([PD.sub.i,t], [FN.sub.i,t], [TM.sub.i,t],
[P.sub.t], [PCY.sub.t], [SH.sub.t]) (1)
where [TH.sub.i,t] = represents the electricity theft variable,
[PD.sub.i,t] probability of theft detection, [FN.sub.i,t] the fine
recovered from culprits and [TM.sub.i,t], the temperature index. (2)
Electricity price [P.sub.t], load-shedding [SH.sub.t] and per capita
income [PCY.sub.t] variables are common for all distribution companies.
All the variables are transformed in their natural logarithmic form. The
model specified in Equation (1) is estimated by Fixed Effects Model
using least-square dummy variable (LSDV) and generalised method of
moments (GMM) methods. Furthermore, the models are estimated using the
variables at their levels as well as in their first differences where
individual effects of utilities are removed. However, the results are
more robust for the variables at their levels and for the instruments in
their first differences hence the results are reported for models at
their levels.
4.1. Utility Specific Characteristics
The electricity theft by a consumer essentially bears some risk of
being detected and fined. The probability of detection or conviction is
constructed by taking ratio of theft detection cases in each utility and
total number of consumers in that utility. Theoretically, it is
plausible to assume that annual cumulative number of detections indicate
the higher probability of being detected ([PD.sub.i,t] thus raising the
associated risk for electricity stealing. So electricity theft is
expected to be negatively related with the probability of detection that
leads to lowering of the electricity theft.
The proposition that crime rate responds to corresponding benefits
and risk, usually is called deterrence hypothesis. The econometric
analysis of criminal behaviour generally applies arrest rates and
sanctions imposed as measures of deterrence. People generally respond to
the deterring incentives and that higher fines increase deterrence for
all groups of individuals [Bar-Ilan and Sacerdote (2004)]. With similar
intuition, the number of cases convicted of electricity theft and
penalty imposed in the form of detection bills are electricity theft
deterrent. Hence, we considered the probability of detection as measured
by the amount of fine recovered ([FN.sub.i,t]).
Temperature index ([TM.sub.i,t]) calculates the intensity of cold
and hot weather in area of operation of autility. Per capita electricity
consumption will rise during extreme temperatures and the relative
benefit of electricity theft will become more likely to offset the cost
in terms of risk of detection for a consumer. Thus the temperature index
is assumed to be positively related with the electricity theft. There
may be potential endogeneity between electricity theft ([TH.sub.i,t])
and cases of theft detection ([PD.sub.i,t]). The higher theft rate may
indicate higher detection cases, implying that higher probability of
detection may be induced by electricity theft. The result would be that
the dependent variable will be correlated with error term in the Fixed
Effects and Random Effects models and the least square estimates would
be biased. To handle this issue, Generalised Method of Moments (GMM) is
also applied or model estimation.
4.2. Country Specific Characteristics
For some variables, we do not have the data for each utility or
region, hence we use the common country level data for all distribution
companies. Average electricity price is positively related with the
electricity theft due to higher net payoff from electricity theft in
case of higher prices. In the presence of low probability of detection,
low fines and widespread corruption the consumers become risk neutral
and theory suggests that theft will tend to increase with tariff rate if
offenders are risk neutral. If the system is already exposed to high
rate of electricity theft, an increase in tariff rates may affect
electricity demand and revenue of utilities in two ways. The honest
consumers may cut their consumption of electricity, while the
proportional number of dishonest consumers may increase their
consumption. The result may be higher electricity consumption, higher
bribe earnings for corrupt employees, higher electricity theft and lower
revenues for utilities. It is due to the expectation that if the tariff
rate is high, it will induce temptation among the consumers to steal
electricity as in this case expected gains would be higher.
The quality of electricity supply service proxied by amount of
load-shedding ([SH.sub.t]) is another interesting variable in our model.
The electricity shortage extensively affects those utilities that have
higher level of theft. On one hand, the higher rate of load-shedding may
reduce total electricity consumption and thus lower the amount of
electricity theft. On the other hand, it may damage the relationship
between the consumers and utility and generate a disregard of peak load
by consumes thus resulting in inefficient use of energy. Thus
load-shedding may increase or decrease electricity theft depending on
the time and duration of load shedding. The rise in per capita income
([PCY.sub.t]) is expected to lower the electricity theft. In general,
the higher income may lead the consumers to avoid risk. Thus the income
is expected to be negatively related with electricity theft.
4.3. Data Description and Sources
The data used in this study consist of a balanced panel from 9
Pakistani distribution companies for the period 1988-2010. The data
mainly obtained from various organisations and publications that mainly
include, Electricity Marketing Data by NTDCL, Planning and Statistics
Departments of WAPDA, Pakistan Meteorological Department, the Federal
Bureau of Statistics and Annual Report of KESC. We employed a number of
company specific variables as well as macroeconomic variables. Table 3
gives the description and sources of data. Electricity theft is our
dependent variable proxied by the distribution losses of electricity
distribution companies in Pakistan. (3) Electricity price is important
in explaining electricity theft and we use average price per unit
(kilowatt hour) obtained by dividing the total revenue from electricity
sale in the country by the electricity supplied.
Currently, National Electric Power Regulatory Authority (NEPRA)
announced a uniform electricity tariff rate in Pakistan and the data for
average sale price at company level is not available, hence we use
electricity price for KESC while all other distribution companies share
the same electricity price. (4) The temperature variable is constructed
by taking sum of degrees above 24 and below 12 from average monthly
temperature at each weather station as follows. The heating degrees (HD)
that require heating the space and water are calculated as follows:
HD=[[summation].sub.j]=1. H(12-[T.sub.j,avg]) ... ... ... ... ...
(2)
where H is a dummy variable equal to 1 if average monthly
temperature at a weather station is below 12[degrees]C, and zero
otherwise. The average monthly temperature in the jth month is
represented by [T.sub.j,avg]. Similarly, the cooling degrees (CD) that
require cooling the space and water are calculated as follows:
CD = [[summation].sub.j=I] x H([T.sub.j,avg.sup.-] 24) ... ... ...
... ... (3)
where C is a dummy variable equal to 1, if average monthly
temperature is above 24[degrees]C. The temperature variable
([TM.sub.t]), defined as a sum of degrees showing extreme temperatures
in a year, is obtained by adding the two measures in Equations (2) and
(3):
[TM.sub.t] = HD + CD ... ... ... ... ... (4)
The temperature variable is obtained by adding monthly
discrepancies in degrees from lower and upper benchmarks at a weather
station. The variable to capture the probability of detection is
constructed by taking the annual number of thefts detections divided by
total consumers for each distribution company.
5. ECONOMETRIC METHODOLOGY
We estimate the fixed effect model by relaxing the restriction on
intercept and let the intercept to vary for each utility, still assuming
that the slope coefficients are constant across the utilities. This is
done in Fixed Effects model due to the tact that the intercept is time
invariant although it varies across utilities. To estimate the Fixed
Effects model, we apply least squares with dummy variables (LSDV)
approach by including the cross-sectional dummies of utilities. The
model can be written as follows.
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (5)
The subscript i denotes the ith utility (i = 1, ..., N) and the
subscript t denotes the jth year (t = 1, ..., T). The subscript i on the
intercept suggests that the intercepts may take different values across
utilities.
The study also estimate the Fixed effects model through the system
GMM to account for the endogeneity of the lagged dependent variable in
the presence of possible autocorrelation in the random error. The GMM
technique requires the specification of a set of moment conditions that
the model should satisfy. It provides robust estimates in that it does
not require information of the exact distribution of errors. For the GMM
estimators to be identified there must be at least as many instrumental
variables (including an intercept) as there are parameters to be
estimated. GMM estimation accounts for unobserved utility specific
effects, allows for the inclusion of lagged dependent variables as
regressors and controls for endogeneity of all the explanatory variables
by selecting parameter estimates such that the sample correlations
between the instruments and the random errors of the model are close to
zero. Least square estimator can also be viewed as a special case of GMM
estimator, based upon the conditions that each of the right-hand
variables is uncorrelated with the random errors of the equation.
The lagged variable on the right hand-side of the equation makes
the model dynamic and changes the interpretation of the equation
considerably. Without lagged variable, the independent variables produce
observed outcome that is, [TH.sub.i,t] representing the full set of
information. The lagged variable brings in the equation the entire
history of the right hand-side variables such that any measured
influence would be conditional on this history. The general approach to
estimate such models relies on instrumental variables on GMM estimator
[Arellano and Bond (1991); Arellano and Bover (1995)]. This is why, we
also used GMM method that handles the potential endogeneity.
The LSDV estimation approach for the Fixed Effects Model is costly
in terms of degree of freedom loss. Judson and Owen (1999) provide a
guide to choosing appropriate techniques for panels of various
dimensions and find that the LSDV estimator only performs well when the
time dimension of the panel is large and propose that GMM is the best
choice overall.
6. RESULTS AND DISCUSSION
This section presents the empirical findings based on the
analytical framework developed in Section 3 by providing a menu of
models, techniques and regressors. The Hausman test for the fixed and
random effects regressions suggests that Fixed Effects Model is more
appropriate in this case since the joint fixed effect is significant at
5 percent. The test statistic is 2.15 with probability 0.035. Hence, the
Fixed Effects Model would be preferred choice on the basis of the test.
Moreover, the results are more robust when models are estimated using
variables at their levels. In order to take the specific nature of nine
companies into account, we employed the Fixed Effects Model estimated
through least square dummy variable (LSDV) regression model and GMM. In
this study, the Fixed Effects Model is interpreted to mean that the
impact of explanatory variables of the Equation (5) on electricity theft
greatly depends on the utility specific characteristics. The results are
presented at Table 4.
The intercept values of the nine utilities are different with
highest in KESC. PESCO stands second followed by QESCO. These
differences are due to the differentials in utility governance and
prevalence of underground economy therein. The Fixed Effects model
estimated with GMM uses the following set of variables as instruments.
List of Instruments:
d([TH.sub.t](-1)) First difference of electricity theft, dependent
variable.
d([PD.sub.t](-1)) First difference of the number of recorded cases
of electricity theft,
d([FN.sub.t](-1)) First difference of the amount of recovery of
line recovered on theft.
d([TM.sub.t](-1)) First difference of the electricity price
variable.
d([SH.sub.t](-1)) First difference of the temperature index.
d([SH.sub.t](-1)) First difference of load-shedding variable.
d([PCC.sub.t](-1)) First difference of per capita electricity
consumption.
d([CPI.sub.t](-1)) First difference of Pakistani score of
corruption perception index taken from Transparency International.
d([EI.sub.t](-1)) First difference of energy intensity by taking
ratio of energy consumption to real GDP.
d([GINI.sub.t](-1)) First difference of Gini coefficient,
indicating income inequality.
d([PCY.sub.t](-1)) First difference of real per capita income.
The results show that model performs well econometrically and the
overall quality of results is satisfactory. The R-square and adjusted
R-square are high enough, indicating strong explanatory power of the
estimated equations. Most of the Durbin-Watson statistics fall in the
non-rejection range indicating absence of considerable autocorrelation.
The significance of t-statistics associated with most of the parameter
estimates further indicates good performance of the estimated models.
The performance of explanatory variables in the model estimated by LSDV
and GMM is discussed in detail below.
The probability of detection variable has poor performance, as
signs of its coefficients are against the theory. The result indicates
that the performance of punishment for conviction or fine remains mixed
in the models. The relatively weak performance of these variables
despite their theoretical relevance to electricity theft may be due to
ineffective surveillance and presence of widespread corruption. The
effect of an increase in electricity price on electricity theft is
positive as expected because rising electricity price increases the
benefit from stealing electricity for the given levels of risk of being
fined. The price variable is found to be significant with highly
significant estimated regression coefficient value in all the models,
signifying the role of electricity tariff rate in explaining electricity
theft in our models. The effect of increase in per capita income on
electricity theft is negative, complying with the assertion that the
individuals become more risk averse as income rises for the same amount
of pecuniary benefit. The per capita income variable significantly
affects the electricity theft with highly significant estimated
coefficient in all the models.
Our findings are consistent with Bo and Rossi (2007). Thus, firms
in those countries would appear to be less efficient, because part of
the energy they effectively distribute gets stolen, rather than sold. It
again indicates the importance of economic variables such as, income and
price and both the variables can be appropriately used for a better
management of the sector in the country. It also shows that in an
electricity supply system burdened with huge losses, an increase in
electricity tariff rate may not increase the revenues of utility as it
may lead to an increased level of electricity theft.
The effect of temperature on energy consumption is well established
and a number of studies have shown that energy consumption is elastic to
extreme temperatures. Table 4 shows that temperature appears significant
with sufficiently high positive coefficient in all the estimated models.
Another variable considered in the models is load-shedding, which has
taken quite low and positive though highly significant coefficient value
in both the estimations suggesting that the deteriorating quality of
service adds to electricity theft.
7. CONCLUSION
Electricity theft is common crime in many countries and electric
utilities worldwide have to forego huge amounts of revenues every year
due to theft of electricity. It causes huge financial losses to
utilities and hurts future investment for capacity additions.
Electricity distribution companies and governments resort to technical
and legal measures to combat this non-violent offense. As a result,
formal laws and technical measures are generally introduced. Rather than
concentrating only on the technical measures and law enforcement, this
study intends to indicate the economic, social and meteorological
factors affecting electricity theft in the context of a developing
country where electricity theft situation is a serious phenomenon.
This paper has empirically investigated the effects of various
factors in explaining electricity theft from electricity distribution
systems using the panel data from nine electricity distribution
companies of Pakistan for the period 1988-2010. The study estimates the
Fixed Effects models using the OLS and GMM techniques. The empirical
evidence from the estimated econometric models is by-and-large
consistent with the conceptual framework, although the impact of the
number of conviction cases is unclear because it either appears with
wrong sign or is statistically insignificant.
The results indicate that the economic factors such as per capita
income of the consumers and consumer price of electricity are key
determinants of electricity theft as suggested by all the models. The
electricity theft is negatively related with per capita income, implying
that an increase in income level lowers the electricity theft with
sufficiently higher coefficient value. The opposite is true for
electricity price, which positively affects the electricity theft. It
also emphasises the importance of minimising electricity theft since in
the presence of widespread theft, the income and price elasticity
estimates for electricity demand cannot be used as policy tools for
achieving electricity conservation and efficiency goals. The effect of
temperature on electricity theft is positive, which seems reasonable as
the extreme temperatures lead to higher electricity consumption that may
consequently induce electricity theft.
The results show that the tariff policy and the overall electricity
demand in the country are important policy variables and the regulatory
body needs to keep these factors in mind in decision-making regarding
the overall electricity supply and tariff late. The results from this
study suggest that electricity price may not be used as an effective
energy conservation tool in the presence of widespread electricity
theft. Moreover, in such cases, excessive demand and power shortfalls
cannot be reduced. The electricity price in Pakistan is already too high
in relation to the quality of service and in real terms. For example,
hours of work to buy 100 units of electricity in Pakistan would be more
than 10 times the hours required to buy the same amount in a country
like the USA. So, hard-core pricing mechanism cannot be applied to many
such countries and the shortfall has to be met in long run through
better planning and management. The equitable electricity prices can be
achieved by minimising the cost of generation. Reduced load-shedding
signify better quality of service that gives a positive gesture to the
consumers, which may in turn oblige them to pay for the service. This
suggests that the issues in supply and demand for electricity are
inter-twined. The findings of the empirical study may be applicable in
most of developing countries where hefty amounts of revenues are lost
due to electricity theft every year.
The study suggests that the issues in supply and demand of
electricity are intertwined. The supply issues can be handled by keeping
the consumer price of electricity right. On one hand, it is inevitable
that utility revenues cover the generation and supply costs for proper
functioning of utilities and sustainable electricity industry.
Increasing electricity prices is a difficult decision for a political
government and the government provides subsidy to electricity in the
short term in view of rising costs of generation. The least cost
optimisation for future electricity generation plans is very important
to avoid price hikes since electricity availability is useless if it is
not affordable. It will induce electricity theft as per analysis.
REFERENCES
Alam, M. S., E. Kabir, M. M. Rehman, and M. A. K. Chowdhury (2004)
Power Sector Reform in Bangladesh: Electricity Distribution System.
Energy 29, 1773-1783.
Arellano, M. and S. Bond (1991) Some Tests of Specification for
Panel Data: Monte Carlo Evidence and an Application to Employment
Equations. Review of Economic Studies 58,277-297.
Arellano, M. and O. Bover (1995) Another Look at the Instrumental
Variable Estimation of Error-components Models. Journal of Econometrics
68:1, 29-51.
Auffhammer, M., C. Blumstein, and M. Fowlie (2008) Demand-side
Management and Energy Efficiency Revisited. The Energy Journal 29:3,
91-104.
Baltagi, B. H. (2005) Econometric Analysis of Panel Data. (3rd
Edition). John Wiley and Sons Ltd.
Bar-Ilan, A. and B. Sacerdote (2004) The Response of Criminals and
Non-criminals to Fines. Journal of Law and Economics 47, 1-17.
Becker, G. S. (1968) Crime and Punishment: An Economic Approach.
Journal of Political Economy 16, 169-217.
Becker, G. S. and G. J. Stigler (1974) Law Enforcement, Malfeasance
and Compensation of Enforcers. Journal of Legal Studies 3, 1-18.
Bo, E. D. and M. A. Rossi (2007) Corruption and Inefficiency:
Theory and Evidence from Electric Utilities. Journal of Public Economics
91:5-6, 939-62.
Clarke, G. and L. C. Xu (2004) Privatisation, Competition and
Corruption: How Characteristics of Bribe Takers and Payers Affect Bribe
Payments to Utilities. Journal of Public Economics 88:9-10,2067-2097.
Estache, A., A. Goicoechea, and L. Trujillo (2006) Utilities
Reforms and Corruption in Developing Countries. (World Bank Policy
Research Working Paper 4081, December).
Eviews 5.0 (2004) Eviews 5.0 User Guide. Quantitative Micro
Software, LLC, 4521 Campus Drive, #336, Irvine CA, 92612-2699.
FODP (2010) Integrated Energy Sector Recovery Report and Plan.
Report by Friends of Democratic Pakistan, Energy' Sector Task
Force, October. Funded by USAID, March.
Greene, W. B. (2003) Econometric Analysis. (5th Edition). Pearson
Education Inc., Singapore.
Gujrati, D. N. (2003) Basic Econometrics. (4th ed). McGraw
Hill-Publishing, Boston.
Gulati, M. and M. Y. Rao (2007) The Many Faces of Corruption:
Tracking Vulnerabilities at the Sector Level, ed. by J. E. Campos and S.
Pradhan. pp. 115-157. (The World Bank, Washington, D.C., USA).
Gupta, S., H. Davoodi, and R. Alonso-Terme (2002) Does Corruption
Affect Income Inequality and Poverty? Economics of Governance 3, 23-45.
Hausman, J. (1978) Specification Tests in Econometrics.
Econometrica 46, 1251-1271.
International Country Risk Guide (1CRG) (2007) The PRS Group. New
York.
IPP (2009) State of the Economy: Emerging from the Crisis. 2nd
Annual Report of Institute of Public Policy (IPP), Beaconhouse National
University, Lahore.
Jamil, F. (2013) On the Electricity Shortage, Price and Electricity
Theft Nexus. Energy Policy 54, 267-272.
Jamil, F. and E. Ahmad (2011) Income and Price Elasticities of
Electricity Demand: Aggregate and Sector-Wise Analyses. Energy Policy
39:9, 5519-5527.
Joseph, K. L. (2010) The Politics of Power: Electricity Reform in
India. Energy Policy 38, 503-51 1.
Judson, R. A. and A. L. Owen (1999) Estimating Dynamic Panel
Models: A Guide For Macroeconomists. Economics Letters 65:1, 9-15.
Katiyar, S. K. (2005) Political Economy of Electricity Theft in
Rural Areas: A Case Study from Rajasthan. Economic and Political Weekly
40:7, 644-648.
KESC (2006) 94th Annual Report. The Karachi Electric Supply
Corporation Limited, Karachi.
Kessides, I. N. (2013) Chaos in Power: Pakistan's Electricity
Crisis. Energy Policy 55, 271-285.
Loughran, D. and J. Kulick (2004) Demand Side Management and Energy
Efficiency in the United States. The Energy Journal 25:1, 19-41.
Lovei, L. and A. McKechnie (2000) The Cost of Corruption for the
Poor--the Energy Case Note No. 207. The World Bank Group-Private Sector
and Infrastructure Network.
Nagayama, H. (2010) Impacts on Investments, and
Transmission/Distribution Loss through Power Sector Reforms. Energy
Policy 38, 3453-3467.
Nakano, M. and S. Managi (2008) Regulatory Reforms and
Productivity: An Empirical Analysis of the Japanese Electricity
Industry. Energy Policy 36, 201-209.
NTDCL (2011) Energy Marketing Data. (35th Ed.) National
Transmission and Dispatch Company Ltd., WAPDA House, Lahore.
Pacudan, R. and E. De Guzman (2002) Impact of Energy Efficiency
Policy to Productive Efficiency of Electricity Distribution Industry in
the Philippines. Energy Economics 24:1, 41-54.
Planning Commission (2013) The Causes and Impacts of Power Sector
Circular Debt in Pakistan.
Smith, T. B. (2004) Electricity Theft: A Comparative Analysis.
Energy Policy 32:18, 2067-2076.
The Electricity Act-1910: with Electricity Rules 1937. (2006) (2nd
ed.) Lahore: Babar Saber Printers.
Transparency International (2007) Corruption Perception Index (CPI)
Various Years Data. Berlin: Web:
http://www.transparency.org/policy_research/surveys_indices/cpi.
Wooldridge, J. M. (2009) Introductory Econometrics: A Modern
Approach. (4th ed.) South-Western Cengage Learning, Mason OH. USA.
(1) Electricity demand is price elastic in case of Pakistan [see,
for instance, Jamil and Ahmad (2011)]. Electricity theft is a criminal
offence subjecting a person to a prison sentence up to three years or
fine up to Rs 5000 or both as per legal provisions of utilities in
Pakistan. See, for example, Electricity Rules 1937. Usually detection
bills may be charged due to the provisions of Section 26A, S-39, S-39-A,
S-44, S-48 on detection of theft or illegal abstraction of electricity
(Electricity Ad-1910).
(2) We tried a number of variables as regressors in the analysis
that appear insignificant including; country level corruption perception
index, Gini coefficient to incorporate income inequality, socioeconomic
index, per capita electricity consumption in each utility, time series
of energy intensity constructed by taking the ratio of energy
consumption in British Thermal Unit (BTU) and real GDP.
(3) The distribution losses include mainly electricity theft and a
small fraction of technical losses [Alam, el at. (2004)].
(4) Average price of electricity may actually vary in different
companies due to varying composition of consumer categories and cross
subsidisation across sectors.
Faisal Jamil <fsljml@hotmail.com> is Assistant Professor,
National University of Sciences and Technology, Islamabad Eatzaz Ahmad
<eatzaz@qau.edu.pk> is Professor and Dean Faculty of Social
Sciences, School of Economics, Quaid-i-Azam University, Islamabad.
Table 1
Profile of the Utilities and Distribution Losses in
Pakistan During 2010
Utility/ Number of Units Units
Distribution Consumers Supplied Billed
Company (Million) (GWh) (GWh)
LESCO 3.18 16,101 13,880
GEPCO 2.45 6,987 6,220
FESCO 2.88 9,329 8,317
IESCO 2.06 8,396 7,572
MEPCO 4.06 12,225 9,915
PESCO 2.94 12,638 8,258
HESCO 1.51 8,275 5,395
QESCO 0.49 5,167 4,099
KESC 2.05 13,362 9,905
Pakistan 17.8 92,480 73,561
Utility/ Distribution Billing
Distribution Losses Recovery
Company (Percent) (Percent)
LESCO 13.7 93
GEPCO 11.0 96
FESCO 10.9 97
IESCO 9.8 96
MEPCO 18.9 94
PESCO 37.0 79
HESCO 34.8 60
QESCO 20.7 76
KESC 34.9 100
Pakistan 20.4 89
Note: GWh=Giga watt hours equivalent to one million KiloWatt
hours, Source: Electricity Marketing Data, 35th Ed.
Table 2
Theft Detection, Penalty and Enforcement in 2009 in Pakistan
Cases Amount of Fine Recovery Percentage
Utility Detected (Rs. Mn) (Rs. Mn) Recovery
LESCO 35,132 320 91 28
GEPCO 34,751 121 94 74
FESCO 36,473 177 94 53
1ESCO 10,700 81 18 22
MEPCO 68,603 315 91 29
PESCO 270,000 1,865 11 0.01
HESCO 376,000 1,505 343 23
QESCO 8,857 16 11 70
KESC 10,700 81 18 22
Source: Statistics Department, WAPDA House, WAPDA Lahore,
and Commercial Wing, KESC.
* Detection Bills are charged on detection of electricity theft that
presumably contain electricity charges plus fine or penalty.
Table 3
Variables and Data Sources
Variable Symbol Variable Definition
Per Capita [PCY.sub.t] Real GDP per capita
Income (Country level data)
Electricity [P.sub.t] Average electricity price
Price (Country level data)
Electricity [TH.sub.t] Distribution losses of electricity in
Theft percent
Probability [PD.sub.t] Number of detection bills divided by
of Detection total number of consumers
Fine per [FN.sub.t] Amount of fines recovered divided
Incidence by number of detection bills (Rs.
Mn)
Load-shedding [SH.sub.t] Percent capacity shortfall of real time
electricity demand (country level
data)
Temperature [TM.sub.t] Population weighted temperature
index of the utilities' regions
Variable Source
Per Capita Federal Bureau of
Income Statistics, Islamabad,
Pakistan
Electricity Planning Department,
Price WAPDA, Lahore
Electricity Electricity Marketing
Theft Data,
NTDCL, Lahore
Probability Statistics Department,
of Detection WAPDA, Lahore
Fine per Statistics Department,
Incidence WAPDA, Lahore
Load-shedding Electricity Marketing
Data,
NTDCL
Temperature Pakistan Meteorological
Department, Islamabad
Table 4
Parameter Estimates of Electricity Theft Models
Variable FE Model LSDV FE Model GMM
Constant 0.196 (c) 0.603 (c)
(1.89) (1.72)
[PD.sub.t] 0.010 (a) 0.013
(5.11) (1.01)
FNt 0.003 (b) -0.004
(2.09) (-0.19)
Pt 0.079 (a) 0.114 (b)
(3.56) (2.37)
TMt 0.037 (b) 0.072 (a)
(2.86) (4.57)
PCYt -0.08 l (a) -0.154 (b)
(-3.41) (-3.02)
SHt 0.008 (a) 0.007 (b)
(4.01) (2.87)
TH(-l) 0.010 (a) 0.009 (a)
(31.83) (7.69)
Fixed Effects
GEPCO 0.016 (a) 0.037 (a)
(3.49) (5.01)
HESCO 0.023 (a) 0.009
(3.58) (0.17)
IESCO 0.019 (a) 0.048 (a)
(4.03) (3.49)
KESC 0.069 (a) 0.071 (c)
(5.76) (1.66)
LESCO 0.008 (c) 0.015
(1.84) (0.52)
MEPCO 0.007 -0.016 (b)
(0.72) (-2.76)
PESCO 0.043 (c) 0.052 (b)
(4.61) (2.34)
QESCO 0.026 (b) 0.049 (b)
(2.48) (2.65)
R-Square 0.94 0.91
Adj. R-Square 0.92 0.90
DW Statistics 1.79 1.71
J-Stat -- 4.82
F-Stat * 10.12 7.89
(Probability) (0.000) (0.000)
Notes: FE stands for Fixed Effects model.
The figures in () represent t-Statistics and superscript a,
b and c denotes the level of significance at 1 percent, 5
percent and 10 percent respectively.
* Wald test of Normalised Restriction (=0), the significance
of dummy variables.
