Energy Policy 39 (2011) 7112–7120

Contents lists available at SciVerse ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

Performance evaluation and improvement directions for

an Indian electric utility
Vinod Kumar Yadav a,n, N.P. Padhy b, H.O. Gupta b
a
Electrical Engineering Department, Gautam Buddha University, Greater Noida 201310, India
b
Electrical Engineering Department, Indian Institute of Technology, Roorkee 247667, India

a r t i c l e i n f o a b s t r a c t

Article history: This study evaluates the performance of 29 Electricity Distribution Divisions (EDDs) of an Indian state –
Received 10 July 2010 Uttarakhand – deploying Input oriented Data Envelopment Analysis (DEA). The results indicate that the
Accepted 16 August 2011 performance of several EDDs is sub-optimal, suggesting the potential for cost reductions and possible
Available online 28 September 2011
reduction in employees number. In the DEA method more than one unit are identified as efficient.
Keywords: Therefore, this study suggests a method for ranking the efficient units by their importance as
Data envelopment analysis benchmarks for the inefficient units through benchmark share measure. The bigger the benchmark
Technical efficiency share, the more important an efficient division is in benchmarking for inefficient ones. Result reveals
Benchmark-share measure that plain area divisions are relatively efficient and have higher potential to influence the performance
of inefficient EDDs. This study is envisaged to be instrumental to policy makers and managers to
increase the operational efficiency of inefficient EDDs and thereby increase the competitiveness in the
face of restructuring and liberalization of Indian electricity sector.
& 2011 Published by Elsevier Ltd.

1. Introduction
In the scenario of ever increasing energy demand, it has been an
anathema for the power utilities of developing nations to operate and
manage their power industries in an efficient and economic way. In
recent years, the power industry privatization and restructuring are
gaining international importance as a measure to achieve this. The
reform process has been initiated in developing countries to techni-
cally and economically inefficient electric utilities and weaker institu-
tions where the sector faced resource crunch (Jamasb, 2006; Singh,
2010). India is no exception to this power sector scenario.
The era of reform and restructuring of Indian power sector was
commenced in 1991 through opening the sector for Independent
Power Producers (IPPs). It was required to improve the dilapidated
financial condition and poor technical and commercial efficiency
of the monolithic State Electric Boards (SEBs) (Thakur et al., 2006;
Totare and Pandit, 2010). Due to inadequate financing, insufficient
revenues and ineffective management, it was not being able to
respond to the energy demand created by economic growth of the
country, though the installed capacity scaled from 1362 MW in
1947 to about 159,398 MW as on march 2010, estimated energy
and peaking shortage in the country for 2010–2011 is 10.6% and
12.1%, respectively (CEA, 2010). To turn around this situation,
n
Corresponding author. Tel.: þ91 9411487447; fax: þ91 120 2344492.
E-mail address: vinod@gbu.ac.in (V.K. Yadav).
the government of India has enacted a number of laws, the latest
being the Electricity Act, 2003 (with amendments in 2004 and
2007) (Tankha et al., 2010).
Although electricity generation has increased many folds over
the previous decades, more than half of the population still does not
have access to electricity (Ghosh, 2010). On top of it the Aggregate
Technical and Commercial (AT&C) losses gnaw about 35% of energy
produced. Moreover, distribution losses in India are also substan-
tially higher (30%) than the global benchmarks. This has resulted
in per capita energy of merely 631 kWh available for consumptions,
whereas the world average stands at about 2873 kWh in
2005–2006 (MoP, 2009). Power sector continued to render unsatis-
factory performance as their main focus was on generation expan-
sion programs only. Whereas the reform in distribution sector should
have also been given an equal or more importance as efficiency
improvement measures, keeping in view that the Indian power
utilities feed a very large number of consumers, located over a wide
area of the subcontinent (Sharma et al., 2005). In order to achieve
commercial viability and to reduce AT&C losses of electric utilities,
ministry of powers formulated a six level intervention strategy and
to operationalize the strategy, ministry launched Accelerated Power
Development Program in the year 2000–2001 (GoI, 2006).
Though efficiency measurement is a time tested and integral
part of power and energy sector in the present era of restructuring
and privatization of electric utilities around the world, productivity
assessment is being viewed with importance. Performance evalua-
tion of an organization highlights the areas and opportunities for

0301-4215/$ - see front matter & 2011 Published by Elsevier Ltd.

doi:10.1016/j.enpol.2011.08.028
 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120
improvement. Organizations of similar nature of activity can
quantify their relative performance comparing the results and
develop strategic plans for improvements in their performance
taking the best in the class as benchmark (Pahwa et al., 2002).
Various research reports on organizational performances have
forced the decision makers to be more concerned about finding
ways to improve productivity. In this regard, Data Envelopment
Analysis (DEA) is probably the most widely used mathematical
approach for benchmarking of organizational units. This technique
measures the relative efficiencies of decision making units (DMUs)
that use inputs and outputs, which are incommensurate in nature.
DEA has been developed by Charnes et al. (1978) and Banker et al.
(1984) based on Farrell’s (1957) pioneering work. Presently it has
grown to be an indispensable tool in the areas of operations
research and management science. It has been applied to a wide
range of managerial and economic problems both in public and
private sectors (Li and Reeves, 1999).
The main objective of the present study is to examine the
performance of Electricity Distribution Divisions (EDDs) of an
Indian State namely Uttarakhand at micro-level and bring forth
the scopes for improvement through DEA application. A bench-
mark-share measure is also developed for technically efficient
EDDs in order to further characterize the performance of efficient
ones (Zhu, 2000; Yang and Lu, 2006) and to yield information on
the role of each efficient division played in benchmarking ineffi-
cient divisions and also to identify the best EDDs in terms of the
benchmark-share.
This paper is organized as follows. After the introductory
section performance of State Owned Electric Utilities (SOEUs) is
discussed in Section 2. Methodology for the DEA analysis and
benchmark-share measure has been discussed in Section 3.
Section 4 discusses the selection of input and output factors.
Section 5 includes results, discussion and policy implication.
Section 6 concludes with the findings of this paper.
2. Performance of state owned electric utilities (SOEUs)
Electricity demand in India has increased significantly over the
past two decades owing to the rapid economic growth in the post
liberalization. High energy shortage and peak deficit are by and
large a common phenomenon in all the states. This can be
attributed to operational inefficiency, non-rationality at all phases
of trading of energy viz. tariff setting, metering, billing, revenue,
etc. and inadequate resources for capacity addition (Sharma et al.,
2005). The generation expansion programs during past decade
suffered substantial slippage from targets. Capacity addition as a
percentage of targets during the 8th, 9th and 10th five year plans
of 54%, 47% and 75%, respectively. However, the shortfall in the
350000
Commercial Losses (Millions of Rs)
Commercial Losses
300000
250000
200000
150000
100000
50000
0 0
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Fig. 1. Performance of SEBs.
(Source: MoP, 2008; Planning Commission, 2002; GoI, 2008).
7113
10th plan has been partly offset by improved plant performance
and reduction of Transmission and Distribution (T&D) losses (IIPA,
2006). In order to meet the growing demand in the coming years,
during the 11th plan, another 62,475 MW of generation capacity
will have to be added. As per the estimation of the Central
Electricity Authority (CEA), this will have to include 13,500 MW
from the private sector. Capacity addition of such a magnitude
will call for an additional investment of approximately 5,000,000
Rs. Million, to be sourced from both the public and private sector
(IIPA, 2006).
In the year 2008–2009, the overall Plant Load Factor (PLF) of
power station is found to be 77.2% with the 84.3% central owned
stations, 71.2% state owned stations and 91% private owned
thermal stations. Plant load factors of thermal power stations have
reflected the growth of only 6% from 1997–1998 to 2008–2009.
Total increase in PLF is mainly contributed by private and central
owned thermal stations and the state owned thermal stations still
delivering unsatisfactory performance (MoP, 2009). Plant avail-
ability has merely increased from 77.8% in 1995–1996 to 81.78% in
2005–2006. Auxiliary consumptions of power plants are still high
at about 8.44% for the same year (IIPA, 2006). Apart from opera-
tional inefficiency, reflected in poor energy efficiency and PLF of
generating plants of SOEUs, the T&D system poses the most serious
burden on the performance of the Indian power sector (Singh,
2006). The Indian power sector is poorly placed in terms of the
level of T&D losses, which have swelled to enormous proportions.
T&D losses were 24.79% in the year 1997–1998 and increased to
28.65% in the year 2006–2007 (MoP, 2008; Planning Commission,
2002). The per capita consumption of energy is amongst the lowest
in the world and the growth rate achieved during last decade is not
appreciable. Even after 6 years of enactment of APDRP, AT&C losses
of state power utilities have not shown much improvement over
the past years. It showed slight decrease from 38.86% to 33.07%
during 2001–2006 (GoI, 2006; MoP, 2009). AT&C losses for Delhi,
Haryana, Maharashtra, Orissa and Rajasthan are between 30% and
40% and for Madhya Pradesh, Uttar Pradesh and Uttarakhand it is
above 40% (PFCL, 2008). It will not be possible to reduce the
technical as well as commercial losses without metering each and
every consumer, smaller or bigger, including agriculture. The
billing efficiency at the national level has increased only from
68.37% during 2002–2003 to 71.04% during 2006–2007. The
national average collection efficiency has merely increased from
92.91% during 2002–2003 to 94.20% during 2006–2007 (GoI, 2006;
MoP, 2009). T&D and commercial losses of SEBs for the period
1997–2006 are presented in Fig. 1.
Financial performance of SOEUs has also been a major area of
concern. The gap between Average Cost of Supply (ACS) and
Average Revenue Realized (ARR) was 49 paise/kWh in 2006–
2007. The average unit cost of supply during 2006–2007 was
40
T&D Losses
35
30
T&D Losses (%)
25
20
15
10
5
7114 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120
Rs. 2.76/kWh whereas average revenue realized was only
Rs. 2.27/kWh (PFCL, 2008). Commercial losses (without subsidy)
were estimated 257,014 Rs. Million in 2007–2008 as compared to
139,630 Rs. Million in 1997. Gross subsidy on sale of electricity,
which was 355,390 Rs. Million in 2005–2006 was estimated to
rise to 431,326 Rs. Million in 2007–2008 (GoI, 2008). The financial
disorders in the SEBs could be attributed to a number of factors,
including high T&D losses, overstaffing, large unpaid bills, poli-
tical interference, etc. Rate of the Return (ROR) on the assets
employed by the SEBs has been turning more and more negative
because of the growing levels of commercial losses. ROR of the
SEBs was deteriorated and have increased from  12.70% in the
year 1991–1992 to  25.12% in the year 2006–2007 (GoI, 2008;
IMaCS, 2007; Planning Commission, 2002).
It is evident from the above portrayal of financial and technical
indicators that a micro-level appraisal of the SOEUs is necessary
in order to identify the causes underlying inefficiencies and the
extent of scope for improvement of these utilities. In this paper
Uttarakhand is selected for the case study because of convenience
in data collection and related enquiries. Main shortfalls related to
the UPCL (2007–2008) are: high AT&C and T&D losses, low billing
(61.04%) and collection efficiency (93.61%) and low level of
household electrification (53%) (PFCL, 2006). This type of analysis
can also be conducted for other states in future.
3. Methodology
3.1. Data envelopment analysis
DEA is a multifactor productivity analysis model for measuring
the relative efficiencies of a homogenous set of DMUs. Charnes
et al. (1978) proposed CCR model that assumes a constant return
to scale (CRS) assumption. The solution of DEA requires that the
weights for inputs and outputs of each DMU be selected to
maximize its efficiency under certain constraints. In other words,
it allows each unit to pick most favorable weights for its specific
situation. Thus, in mathematical terms the CCR model is given as
Ps
vr yro
Max yo ¼ Prm¼ 1
i ¼ 1 ui xio
s:t:
Ps
vr yrj
0 r Prm¼ 1 r1; 8j
i ¼ 1 ui xij
yo ,vr, Z 0 ,ui Z 0 ; i ¼ 1,2,. . .m; r ¼ 1,2,. . .,s
where y is the relative efficiency score, o is the designated/test
DMU for an optimization run, yrj is the amount of output r
produced by DMU j, xij is the amount of input i utilized by DMU
j, vr is the weight given to the output r, ui is the weight given to
the input i, n is the total number of DMUs, m is the total number
of inputs and s is the total number of outputs.
The fractional program in (1) is subsequently converted to a
linear programming format and a mathematical dual is employed
as follows:
Min yo
s:t:
X n
lj xij r y0 xio , i ¼ 1,. . .,m
j¼1
X
n X
lj yrj Zyro , r ¼ 1,. . .,s
j¼1
yo , lj Z 0 8i and r
where lj is the weight in the dual model for the inputs and
outputs of the n DMUs.
In this paper Banker, Charnes and Cooper (BCC) model with
variable return to scale assumption is applied (Banker et al.,
1984). It decomposes the overall efficiency (OE) into technical
efficiency (TE) and scale efficiency (SE). Since model applied is
input oriented, therefore objective of each EDD is to minimize its
inputs, keeping the output level constant.
The technical efficiency of each EDD can be computed as a
solution to the following linear programming (LP) problem with
P
convexity constraint nj¼ 1 lj ¼ 1 added to (2)
Min yo
s:t:
X n
lj xij r y0 xio , i ¼ 1,. . .,m
j¼1
X
n
lj yrj Z yro, r ¼ 1,. . .,s
j¼1
X
n
lj ¼ 1
j¼1
yo , lj Z 0 8i and r ð3Þ
By varying the index ‘o’ over all EDDs, we can get the technical
efficiency (TE) for each EDD. Scale efficiency is the ratio of OE/TE. If
TE¼1, then the EDDo is technically efficient, i.e. lie on the efficiency
frontier. And if TEo1, then EDDo is technically inefficient.
The value of lj obtained by the solution indicates whether the
jth EDD serves as a role model or benchmark for EDDo. If lj ¼ 0,
then the jth EDD is not a benchmark. However, if lj 4 0, for
example lj ¼ 0:6, then the jth EDD is a benchmark EDD with a 60%
weight placed on having potential to derive the target efficient
output and input levels for EDDo. And if lj ¼ 1, then EDDj is
assigned as a self-evaluator, i.e. it acts as a benchmark for itself.
3.2. Benchmarks and benchmark-share
‘‘It is significant to know the extent of inefficient DMUs and
contribution of efficient DMUs towards the inefficient one. Sev-
eral measures have been proposed by the researchers. Smith and
Mayton (1987) have suggested the counter for a specific DMU act
as a reference. Later, the ranking of efficient DMUs has been
proposed by measuring the radial distance from a frontier con-
structed by the remaining DMUs by Andersen and Petersen
(1993). The measurement of ranking using lambda values in the
VRS model by Torgersen et al. (1996); ranking via benchmark
ð1Þ share measurement by Zhu (2000) and Yang and Lu (2006); super
efficiency model using slack based measure (SBM) by Tone (2002)
are some important contributions in this direction’’.
Here, for a particular inefficient EDDd the factor specific (kth
input specific and qth output specific) measure is determined
through the following two linear programming problems solved
by LP Solver LINDO (2003) and the existing BCC model’s best
practice frontier.
The kth input-specific DEA model is given as
ykdn ¼ Min ykd , d A N
s:t:
X
ldj xkj ¼ ykd xkd , k A f1,. . .,mg
jAE
X
ldj xij r xid , i ak
jAE
ldj yrj Zyrd , r ¼ 1,. . .,s
jAE
X
ð2Þ ldj ¼ 1
jAE
ykd , ldj Z0, j A E ð4Þ
 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120 7115

The qth output-specific DEA model is given as group of parameters suggested in the survey by Jamasb and Pollitt
(2001).
fqd q
n
¼ Max fd , dAN
The following inputs and outputs parameters related to EDDs
s:t: are selected for analysis in the present work:
X
ldj yqj ¼ fqd yqd , q A f1,. . .,sg
jAE Inputs:
X x1: Operating and Maintenance (O&M) cost (Rs. Million); it
ldj yrj Z yrd , raq
includes Repairing and Maintenance (R&M), Administrative
jAE
X and General (A&G) costs.
ldj xij rxid , i ¼ 1,. . .,m x2: Number of employees; it includes technical, non-technical
jAE
as well as administrative employees.
X
ldj ¼ 1 Outputs:
jAE y1: Energy sold (Million Units); it represents the energy sold
annually.
fqd , ldj Z0, j A E ð5Þ
y2: Number of customers; it is the sum of connection points,
ldj ykd ldj fqd are
n n n n
where and are optimal values in (4) and and which is supplied to customers.
optimal values (5). y3: Duration of interruption per feeder (Hours); it includes
Here, E and N, respectively, represent the index sets for the shut down, break down and roastering type of interruptions.
efficient and inefficient EDDs identified by the BCC model (3). The y4: Distribution line length (Circuit kilometer); it includes
factor specific measure in (4) and (5) determine the maximum overhead network of 33 kV, 11 kV and LT lines.
potential decrease of an input and increase of an output while y5: Transformer capacity (kVA); transformer capacity is in the
keeping other inputs and outputs at current levels. These factor- range from 16 kVA to 1000 kVA.
specific measures are still multi-factor performance measures, y6: Total sanctioned load per square kilometer (kW/km2); it
since all concerned factors are considered in a single model. includes domestic, commercial, industries, public and private
The kth input-specific benchmark-share measure for each electric systems (irrigation, traction, street lights).
efficient EDD is measured by (6), j A E is
P dn kn Input generally reflects the resources utilized by EDDs. Operating
d A N lj ð1yd Þxkd and maintenance (O&M) cost constitutes one of the important
Dkj ¼ P kn
ð6Þ
d A N ð1yd Þxkd controllable inputs for maintaining the EDDs under specific value.
Apart from this, in our analysis of efficiency, number of employees is
The qth output-specific benchmark-share measure for each
considered as an input, which includes technical, non-technical and
efficient EDD is calculated using (7), j A E
administrative staff. The output reflects the service provided to the
P dn qn
q d A N lj ½1ð1=fd Þyqd
customers. The amount of electricity sold and number of customers
Pj ¼ P qn
ð7Þ are the two important output parameters in electricity distribution,
d A N ½1ð1=fd Þyqd
since each customer has a particular pattern of electricity consump-
The benchmark-share Dkj (or Pqj ) measure the contribution of tion and load characteristic (Salvanes and Tjotta, 1990, 1994). It is
an efficient EDD to the potential input (output) improvement in assumed that administrative employees required for each division is
dn kn dn
inefficient EDDs and depends on the value of lj and yd (or lj proportional to the number of customers and that the technical
kn
and fd ). employees are proportional to the number of customers as well as
The normalized weights are expressed as to the total length of the network and total transformer capacity.
" # Also network length is predominantly the factor of geographical
kn
ð1yd Þxkd
P characteristics and population density of the area served by a
kn
d A N ð1yd Þxkd particular EDD. And transformer capacity is determined by load
forecast characterized by the type of consumers served by the
and
( ) divisions (Miliotis, 1992). Distribution line length and transformer
kn
½1ð1=fd Þyqd capacity are outlets in the process of electrification and industria-
P kn lization, therefore treated as outputs (Yadav et al., 2010). Output is
d A N ½1ð1=fd Þyqd
also characterized by certain quality measures; therefore average
kn dn
Here, (1 yd )xkd
and ½1ð1=fq Þyqd describe the potential duration of interruptions is included in this study as an important
decrease on the kth input and increase on the qth output, controllable factor. State of Uttarakhand has different geographical
P P
respectively, and the value of j A E Dkj ¼ 1 and j A E Pqj ¼ 1: terrain with unevenly distributed population, therefore to identify
Eqs. (6) and (7) show that an efficient EDD will have a zero the effect of operating environmental conditions on the performance
benchmark-share measure if it does not act as a referent EDD for of divisions; total sanctioned load per square kilometer is included
any inefficient EDD. The benchmark-share measure is directly to produce a more realistic result.
proportional to the efficiency of EDD in benchmarking. To substantiate the choice of variables (analytically), as being
conducive to Indian conditions, a statistical verification is carried
out. Table 1 represents the statistical parameters evaluated for
4. Selection of inputs and outputs finding out the relevance of the choice of indicators for accessing
the performance of EDDs.
Selection of input and output is the most important step in the
process of performance evaluation while using DEA. In this paper  The R-squared of the regression is the fraction of the variation
the selection of inputs and outputs are made based on the insight in the dependent variable that is accounted for (or predicted
given by related case study of Miliotis (1992). The model, which is by) the independent variable.
closely related to operating efficiency from four different models  The F ratio is a measure of how much the model has improved
for input–output selection proposed by the author is chosen in the prediction of the outcome compared to the level of
the present work. The choice of variables is also based on the inaccuracy of the model (Thakur et al., 2006).
7116 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120

R2 values in Table 1 indicate that 56% of the variation in O&M
cost and 52.5% of the variation in number of employees are subject
to the six independent variables; energy sold, number of custo-
mers, duration of interruptions, distribution line length, transfor-
mer capacity and total sanctioned load per square kilometer. Cross
validity of this model can be assessed by comparing the adjusted R2
values with that of observed value of R2. Results of computation of
R2 using Steins formula (Field, 2000) to yield adjusted R2, found to
be almost similar. This validates the model to be good enough.
Table 1 also gives the summary of analysis of variance (ANOVA)
result. F value for O&M cost and number of employees observed to
be 4.610 and 4.058, respectively. Both F values are found to be
significant at po0.010, which implies that there is less than 1% of
chance that results would have come up in a random distribution,
and there is 99% probability of being correct that the variables are
having some effects and assumed that the chosen model is correct.
Thus, in the proposed model the variables O&M cost and number of
employees are well explained by the chosen independent variables.
Operational data of UPCL (2007–2008) is recorded at two
levels. Some data viz. O&M cost is recorded at zonal head quarters
while the other data are recorded at state Headquarter. All the
data related to this study for the year 2007–2008 are collected
from two zones (Kumaon and Garhwal) and UPCL Headquarter
(Urjabhawan, Dehradun). Annual data for duration of interrup-
tions was not available, therefore, monthly data is considered for
this study. It is generated by averaging the interruption data of
the month of May and October. An overview of the statistical
Table 1
R2 and ANOVA results.
analysis of the data across circles is presented in Table 2 in the
form of mean, standard deviation, maximum and minimum
values. According to Cooper et al. (2001) number of EDDs should
be least triple the total number of input and output factors
considered, while using the DEA model. In our study the number
of EDDs is 29, which is more than thrice of the selected 8 input
and output factors.
For the validation of DEA model, we examined the assump-
tions of the isotonicity relationships between the input and
output factors i.e., an increase in any input should not result in
a decrease in any output (Golany and Roll, 1989). If the correla-
tion of the selected input and output factors is positive (0–1), the
factors are isotonically related and can be included in the analysis.
And alternatively, when the correlation between input and output
is negative, then the variable should be omitted from DEA
analysis. Inclusion of distribution losses (input) as a performance
indicator would have yielded better efficiency result because it is
indispensable for this kind of analysis. Correlation of distribution
losses with O&M cost, number of employees and distribution line
length is found to be negative; therefore it is omitted from the
analysis. Table 3 lists the correlation coefficients and R2 results
between the selected input and output variables. Upper value
shows the person correlation between the variables and value in
the parenthesis indicates the significance level. It is observed that
O&M cost has the positive correlation with all output variables.
However, it shows only significant correlation with number of
customer and distribution line length. Similarly, number of
employees indicates positive correlation with all output variables.
Nevertheless, it shows only significant correlation with number of
customer, distribution line length and transformer capacity.

Dependent variable R2 Adjusted R2 F-value Significance

5. Results and discussion
O&M cost 0.560 0.440 4.610 0.00
Number of employees 0.525 0.396 4.058 0.007
5.1. DEA results
Independent (predictors): (constant) energy sold, number of customers, duration
of interruptions, distribution line length, transformer capacity and total sanc- Considering the controllability of the parameters the input
tioned load per square kilometer. oriented CCR and BCC model is applied to determine the efficient

Table 2
Summary statistics of the data across circles.

Variables Mean Median St. deviation Minimum Maximum

x1: O&M cost (Rs. Million) 46.75 40.55 14.51 31.08 71.19
x2: number of employees 615 508 227 437 1029
y1: energy sold (Million Unit) 647.52 516.51 432.56 218.33 1493.10
y2: number of customers 175,775 162,259 41,359 132,622 262,256
y3: duration of interruption/feeder (h) 54.55 61.9 23.73 22.7 84.60
y4: distribution line length circuit kilometer 11,863.58 10,338.12 7315.26 2974.03 24,124.06
y5: transformer capacity 4875 5239 2550 1048 8062
y6: total sanctioned load/square kilometer 685.76 534.29 530.87 90.91 1465.67

Table 3
Correlation matrix and significance level.

Variables x1 x2 y1 y2 y3 y4 y5 y6

x1 1
x2 0.40 (0.03)n 1
y1 0.12 (0.53) 0.20 (0.29) 1
y2 0.57 (0.00)nn 0.66 (0.00)nn 0.34 (0.07) 1
y3 0.02 (0.93) 0.06 (0.77) 0.31 (0.11) 0.08 (0.67) 1
y4 0.64 (0.36) 0.53 (0.00)nn 0.05 (0.82) 0.49 (0.01)nn 0.27 (0.16) 1
y5 0.18 (0.36) 0.38 (0.04)n 0.27 (0.16) 0.34 (0.07) 0.44 (0.02)n 0.36 (0.06) 1
y6 0.08 (0.67) 0.02 (0.90) 0.57 (0.00)nn 0.31 (0.10) 0.05 (0.81) 0.29 (0.13) 0.02 (0.93) 1

n
Correlation significant at the 0.05 level.
nn
Correlation significant at the 0.01.
 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120 7117

frontier for 29 Electric Distribution Divisions (EDDs) by using
DEAP Version 2.1, a computer program by Coelli (1996). The
overall efficiency is decomposed into technical efficiency and
scale efficiency and the nature of return to scale is presented in
Table 4. In model 1 the mean overall efficiency score of the
divisions is 0.812, which implies that there is immense room for
efficiency enhancement. Plain area divisions outperform hilly
divisions. Eight Plain area EDDs lies on the frontier while none
of the hilly divisions are identified as overall efficient and their
efficiency scores lies between 0.598 and 0.948.Only 45% of the
total EDDs are technically efficient and the average mean of
technical efficiency is 0.859, which implies that management is
not efficient in resource utilization. Greater inefficiency is pre-
vailing among hilly divisions and only three (Srinagar, Tehri and
Pithoragarh) are found to be technically efficient. Performance of
Rudrapur circle is relatively good as all the divisions are efficient.
As far as return to scale is concerned 31% EDDs are operating at
appropriate scale (constant return to scale) and the rest is
presenting unbalanced status of scale. It can be identified from
Table 4, that the bigger hilly EDDs like Uttarkashi, Tehri and
Pithoragarh are inefficient mainly due to scale inefficiencies. Tehri
and Pithoragarh are technically efficient but scale inefficient and
represented decreasing return to scale, hence it is evident that
their scale size to be decreased for efficiency improvement.
To deal with the fact that the units may not operate under
equivalent environmental terms i.e. the units corresponding to
plain region and higher population density areas being in more
favorable position. DEA results also supported this assumption.
Therefore, to identify the effects of geographical characteristics on
the performance of EDDs, Model 2 is run in which an additional
output total sanctioned load per square kilometer is considered
for the analysis. Table 4 also presents the results of model 2.
Average efficiency score of model 2 increased by 3.7% as com-
pared to model 1. Efficiency of plain and semi-plain area divisions
found to be increased. No increase in efficiency is observed for
hilly divisions. Plain area divisions Dehradun (S), Dehradun (C),
Roorkee (U) turned out to be efficient in model 2. Thus, from
results of model 1 and model 2, it can be concluded that none of
the hilly divisions are benefited by total load sanctioned per
square kilometer factor.
The low value of scale-efficiency and the fact that these
utilities display either decreasing or increasing return to scale

Table 4
The results of CCR and BCC model.

No. Divisions/DMUs Overall Technical Scale Saving Return to scale Overall efficiency Geographical /terrain

Efficiency O&M cost Number of

(Rs. millions) employees

Model-1 Model-2

EDC (R) Dehradun

1 Vikasnagar 0.884 0.885 0.998 0.695 10 IRS 0.884 Semi-plain
2 Rishikesh 0.817 0.836 0.976 1.268 21 DRS 0.825 Semi-plain
3 Dehradun (R) 1.00 1.000 1.000 0.000 0 CRS 1.000 Plain
4 Uttarkashi 0.598 0.633 0.944 3.953 64 DRS 0.598 Hilly

EDC (U) Dehradun

5 Dehradun (N) 0.511 0.515 0.991 5.363 75 IRS 0.825 Semi-plain
6 Dehradun (S) 0.670 0.763 0.878 3.841 34 DRS 1.000 Plain
7 Dehradun (C) 0.725 0.844 0.859 2.425 21 DRS 1.000 Plain

EDC Srinagar
8 Srinagar 0.948 1.000 0.948 0.000 0 DRS 0.948 Hilly
9 Tehri 0.753 1.000 0.753 0.000 0 DRS 0.753 Hilly
10 Gopeshwar 0.737 0.847 0.871 2.234 33 DRS 0.737 Hilly
11 Kotdwar 0.829 1.000 0.829 0.000 0 DRS 0.829 Semi-plain
12 Rudraprayag 0.712 0.793 0.899 2.435 27 DRS 0.712 Hilly

EDC Roorkee
13 Roorkee (U) 0.967 1.000 0.967 0.000 0 DRS 1.000 Plain
14 Roorkee (R) 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain
15 Haridwar (U) 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain
16 Haridwar (R) 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain

EDC Haldwani
17 Nainital 0.772 0.772 1.000 2.222 39 CRS 0.772 Hilly
18 Ramnagar 0.522 0.586 0.890 3.968 41 IRS 0.522 Semi-plain
19 Haldwani (U) 0.541 0.542 0.998 4.677 57 IRS 0.542 Plain
20 Haldwani (R) 0.650 0.666 0.976 3.622 37 IRS 0.752 Semi-plain

EDC Rudrapur
21 Kashipur 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain
22 Bazpur 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain
23 Rudrapur 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain
24 Sitarganj 1.000 1.000 1.000 0.000 0 CRS 1.000 Plain

EDC Ranikhet
25 Ranikhet 0.755 0.782 0.965 2.631 26 DRS 0.755 Hilly
26 Almora 0.892 0.900 0.991 0.712 23 DRS 0.892 Hilly
27 Bageshwar 0.673 0.699 0.963 3.886 34 IRS 0.673 Hilly
28 Pithoragarh 0.749 1.000 0.749 0.000 0 DRS 0.749 Hilly
29 Champawat 0.843 0.844 0.999 1.529 17 IRS 0.843 Hilly

Average 0.812 0.859 0.946 1.567 19 – 0.849 –

EDC: electricity distribution circle; U: urban; R: rural; N: north; S: south; C: central; CRS: constant return to scale; IRS: increasing return to scale; DRS: decreasing return to scale.
7118 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120

indicate that these divisions have considerable scope for improve-
ments by resizing their scales of operations to the optimal value
defined by more productive utilities in the sample. Result implies
that several EDDs are technically inefficient and relative scales of
these divisions have unbalanced status and needs attention for
efficiency improvement.
5.2. Policy implication
In this study BCC formulation of model 1 is applied to quantify
the potential for possible savings in O&M costs and number of
employees (Thakur et al., 2010). Table 4 demonstrate the need for
induction of cost efficiency among the divisions in the power
supply services owned by UPCL; as per the analysis result, it is
theoretically possible to save Rs. 1.567 Million, yearly, invoking
higher efficiency in the current practices of UPCL. The overall cost
savings may yield cash surplus sufficient enough to improve
services to the customers, and to extend the service to the
potential customers. It is worth mentioning that, in Uttarakhand
access to power still remains a cherished goal i.e. power to all,
which is to be achieved by the year 2012. Only 53% of the
households have been electrified so far (PFCL, 2006).
Electrification of the entire country will incur gigantic expen-
diture and vary from division to division. The progressive electri-
fication and accessibility will pose great challenge to the electricity
divisions over the next couple of years both from technical as well
as socio-economic aspects. The efficiency improvement of EDDs
will have considerable contribution on saving and generate liquid-
ity. O&M cost constitutes three major controllable components
that are employee’s costs, repairing and maintenance (R&M) costs
and administrative and general (A&G) costs. Since in this analysis
number of employees has been also taken as an input, therefore
employees costs are excluded from the O&M cost.
A&G expenses constitutes rent, rates, taxes, insurance, tele-
phone, postage and telegrams, legal and professional charges,
audit fees, fees and subscriptions, conveyance and traveling,
electricity and water charges, printing and stationary, advertise-
ment and promotion. A major chunk of these expenses can be
mitigated by keeping updated with the advancement in the field
of information technology. Thus, it is necessary to periodically
review and update the maintenance program using a structured
method.
The possible savings in terms of number of employees reveals
the extent of inefficiencies prevalent with respect to the over-
staffing in UPCL. Theoretically, there is a potential of curtailment
559 of employees at various division level, constituting approxi-
mately 13% of the total current staff strength in all division. It
must be pointed out here that the actual potential of staff in real
condition might be lesser, since the present model may not have
captured the entire variable. Furthermore, practically the require-
ment of manpower may inflate because of expansion of electri-
fication accessibility by virtue of the government’s aim to provide
‘‘power for all’’ by the year 2012. Corrective measures towards
this problem adopted through stopping the further recruitment of
some categories of the fourth grade employees and these employ-
ees are employed on the contract basis as per the need.
Challenges of dealing with political interference and mono-
polistic nature of electricity market are major concerns for the
electricity sector. The efficiency and performance improvement of
divisions can be ensured by proper incentive policy and motiva-
tion for competitive divisions and its workforce. The parameters
of effective operation management can be accuracy of system
operation and/or effectiveness of the internal audit system. The
internal audit can be made effective by taking leverage of
information technology advancement and developing specialized
aptitude using specific procedure and meta-cognitive training.
5.3. Identification of benchmark and benchmark share
Efficient DMUs are designated as benchmarks for inefficient
DMUs. Inefficient DMUs can try to emulate their benchmarks to
improve their efficiency (Ramanathan, 1966).
The benchmark share measure is developed for each frontier
division of model-1 in order to further characterize the perfor-
mance of technically efficient divisions. Table 5 summaries the
benchmark share of the technically efficient EDDs, with the
ranking mentioned in parenthesis and ordered by the average
rank of the efficient units. This analysis gives the ranking list of
the performance model for all the 13 technically efficient EDDs.
With this we can now identify the inputs and outputs that are
most influential in increasing the efficiency; also those divisions
can be now zeroed upon with ease, which is to be treated as
benchmarks, looking at the average ranking.
Technically efficient division Haridwar (R) has highest share
measure in both the inputs O&M cost (x1) and Number of
employees (x2). Roorkee (R) has a leading role in terms of
Transformer capacity (y5) and duration of interruptions per feeder
(y3). As far as distribution line length (y4) is concerned Dehradun
(R) contributed highest share as a benchmark for other inefficient
divisions. Whereas divisions Roorkee (U) and Bazpur occupy the
first rank in terms of benchmark share for outputs number of
customers (y2) and energy sold (y1), respectively. All the above
stated EDDs with the highest share measure are overall efficient
except Roorkee (U) with 96.7%. Thus, it can be said that EDDs that

Table 5
Benchmark share for inefficient EDDs.

EDDs/DMU Input factors Output factors Average rank

x1 (%) x2 (%) y1 (%) y2 (%) y3 (%) y4 (%) y5 (%)

Haridwar (R) 51.00 (1) 44.82 (1) 3.01 (8) 8.37 (5) 23.39 (2) 24.89 (2) 15.01 (2) 3
Dehradun (R) 7.88 (4) 20.57 (2) 3.94 (6) 12.39 (3) 3.86 (5) 53.06 (1) 6.98 (6) 3.8
Srinagar 10.17 (3) 0.00 (11.5) 3.20 (7) 9.91 (4) 3.22 (6) 9.17 (3) 13.98 (3) 5.36
Roorkee (U) 16.18 (2) 7.41 (4) 0.00 (12) 42.86 (1) 17.08 (4) 0.01 (10) 4.82 (5) 5.43
Roorkee (R) 0.94 (8) 2.88 (6) 1.12 (10) 1.86 (7) 30.63 (1) 2.31 (5) 51.15 (1) 5.43
Kashipur 0.91 (9) 0.00 (11.5) 25.15 (2) 0.53 (8) 18.91 (3) 2.21 (6) 2.27 (8) 6.78
Bazpur 2.06 (7) 2.97 (5) 25.61 (1) 0.00 (11.5) 0.80 (9) 4.90 (4) 0.00 (11.5) 7.00
Sitarganj 0.00 (11.5) 18.46 (3) 9.32 (4) 15.72 (2) 0.00 (11.5) 0.00 (12) 0.00 (11.5) 7.93
Rudrapur 3.34 (6) 0.59 (9) 21.33 (3) 0.00 (11.5) 0.00 (11.5) 2.02 (7) 0.00 (11.5) 8.5
Kotdwar 0.00 (11.5) 0.80 (8) 2.30 (9) 8.27 (6) 0.89 (8) 0.00 (12) 2.43 (7) 8.78
Haridwar (U) 7.52 (5) 0.00 (11.5) 5.02 (5) 0.00 (11.5) 0.00 (11.5) 0.00 (12) 0.00 (11.5) 9.71
Tehri 0 (11.5) 0.00 (11.5) 0.00 (12) 0.09 (9) 1.22 (7) 1.28 (8) 0.75 (9) 9.71
Pithoragarh 0 (11.5) 1.50 (7) 0.00 (12) 0.00 (11.5) 0.00 (11.5) 0.15 (9) 2.62 (6) 9.78
SUM 100 100 100 100 100 100 100
 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120
Fig. 2. Share of efficient EDDs in O&M cost for efficiency improvement.
are technically as well as scale efficient have the highest bench-
mark share in input and output factors for inefficient ones. The
bigger the benchmark share, the more important an efficient
division is in benchmarking. Fig. 2 shows graphically in a pie-
diagram the benchmark-shares of technically efficient EDDs that
should be pointed out as examples for inefficient units for
efficiency improvement while considering O&M cost variable.
Haridwar (R) alone refers to half of the potential improvement
in O&M cost. Srinagar and Roorkee (U) have benchmark-share
more than 10%. Particularly if improvement in O&M cost is of
primary importance, these divisions should be studied in order to
learn. Though efficient other divisions in the diagram have only
minor benchmark-shares and therefore either referenced by few
inefficient divisions or by themselves.
Divisions like Kotdwar, Haridwar (U), Tehri and Pithoragarh
are poorly benchmarked by inefficient EDDs with benchmark
share below 10% and for most of the inputs and outputs it is 0%.
Though these EDDs are efficient, they reveal to be very different in
the input output space to be referenced by the inefficient ones
and they are termed as self-evaluators. It can also be noticed from
Table 5 that plain area divisions are dominating in the benchmark
share and hold relatively higher average rank. This indicates that
these divisions are close in input/output space to the inefficient
ones, i.e. efficiently producing a similar product-mix using a
similar mix of resources. Hilly divisions Tehri and Pithoragarh
occupy the lowest rank in the average rank list. This result is quite
realistic since scale of various inputs, e.g., O&M cost and number
of employees is more easily attained for plain efficient EDDs i.e.
these EDDs have higher potential for efficiency improvement of
inefficient units. It is relatively more difficult for inefficient units
to imitate the scale of hilly efficient EDDs for efficiency enhance-
ment. Exception is there with Srinagar though belonging to hilly
area is at third rank, but it does not have the highest share
measure for any input and output factors. The average ranking of
the benchmark share can be treated with paramount importance
as it point out the EDDs that may be referred as benchmarks by
inefficient ones for performance improvement.
6. Conclusion
With the application of DEA, this study explores the perfor-
mance of 29 electricity distribution division of Uttarakhand. Only
eight divisions are observed to be overall efficient. It is found that
divisions suffered from technical as well as scale inefficiency.
Mean technical efficiency score is found to be 85.9%, and only
7119
13 EDDs are technically efficient. Plain area divisions are gener-
ally more effective in integrating their resources and operate
more efficiently. Among the EDDs of hilly area only three are
found to be technically efficient.
Two models i.e. model-1 and model-2 were taken into account
to monitor the effect of different operational environmental
conditions on the performance of divisions. Overall efficiency
mean score of model-2 was found to greater than model-1 and
performance of plain and semi-plain divisions improved in
model-2. No change in efficiency is observed for hilly divisions
in the second model. The results of the study also indicate the
existence of scale inefficiencies, and suggest that resizing the
present scale of operations may help the divisions to reduce their
scale inefficiencies. Ranking of technically efficient EDDs has been
drawn using the benchmark share measure. Plain area divisions
are dominating in the benchmark share and are frequently
referenced by inefficient ones. These benchmarks may be referred
for performance improvement by the inefficient ones. However, it
is suggested that the selection of any EDD as benchmark should
be done by considering the average rank attained by each
technically efficient EDD in the benchmark share measure analy-
sis. A higher rank evidently demonstrates the higher potential to
influence the performance when referenced by the inefficient
ones. The findings of this research work can assist managers to
improve the overall efficiency of the EDDs of UPCL. In this study
the effect of distribution losses has not been considered for the
efficiency evaluation. Inclusion of this vital parameter should
yield realistic outcome for the study of other electric distribution
utilities.
Acknowledgement
The authors would like to thank the generous assistance
provided by the Zonal headquarters (Dehradun and Haldwani)
and State headquarters/Urja Bhawan (Dehradun) of Uttarakhand
Power Corporation Limited for providing the valuable data. More-
over, the authors would like to thank the anonymous reviewers
for their constructive comments.
References
Andersen, P., Petersen, N.C., 1993. A procedure for ranking efficient units in data
envelopment analysis. Management Science 39, 1261–1264.
Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some models for estimating
technical and scale inefficiencies in data envelopment analysis. Management
Science 30, 1078–1092.
7120 V.K. Yadav et al. / Energy Policy 39 (2011) 7112–7120
CEA, 2010. Load Generation Balance Report 2010–11. Central Electricity Authority.
Government of India.
Charnes, A., Cooper, W.W., Rhodes, E., 1978. Measuring the efficiency of decision
making units. European Journal of Operational Research 2, 429–444.
Coelli, T., 1996. A Guide to DEAP Version 2.1: A Data Envelopment Analysis
Program. Website: /http://www.une.edu.au/econometrics/cepa.htmS.
Cooper, W.W., Li, S., Seiford, L.M., Tone, K., Tharall, R.M., Zhu, J., 2001. Sensitivity
and stability analysis in DEA: some recent development. Journal Productivity
Analysis 16, 31–47.
Farrell, M.J., 1957. The measurement of the Productive efficiency. Journal of the
Royal Statistical Society 120 (3), 253–281.
Field, A., 2000. Discovering Statistics using SPSS for Windows: Advanced Techni-
ques for the Beginners. Sage Publication, London, Great Britain.
Ghosh, S., 2010. Status of thermal power generation in India—perspective on
capacity generation and carbon dioxide emission. Energy Policy 38,
6886–6899.
GoI, 2006. Report on Restructuring of APDRP. Government of India, Ministry of
Power Website: /http://www.powermin.nic.inS.
GoI, 2008. Economic Survey (2007–08). Government of India. Website: /http://
indiabudget.nic.inS.
Golany, B., Roll, Y., 1989. An application procedure for DEA. OMEGA 17 (3),
237–250.
IIPA, 2006. Study on Impact of Restructuring of SEBs. Indian Institute of Public
Administration, Government of India.
IMaCS, 2007. Power Market and Opportunities. A Report For Indian Brand Equity
Foundation (IBEF). Website: /http://www.ibef.org/S.
Jamasb, T., 2006. Between the state and markets: electricity sector reform in
developing countries. Utilities Policy 14, 14–30.
Jamasb, T., Pollitt, M., 2001. Benchmarking and regulation: international electricity
experience. Utility Policy 9, 107–130.
Li, X.B., Reeves, G.R., 1999. A multiple criteria approach to data envelopment
analysis. European Journal of Operation Research 115, 507–517.
Miliotis, P.A., 1992. Data envelopment analysis applied to electricity distribution
districts. Journal of the Operational Research Society 43 (5), 549–555.
MoP, 2008. Annual Report (2007–08). Power: The Building Block of the Economy.
Ministry of Power, Government of India, New Delhi. Website: /http://www.
powermin.nic.inS.
MoP, 2009. Annual Report (2008–09). Power for All. Ministry of Power, Govern-
ment of India, New Delhi. Website: /http://www.powermin.nic.inS.
Pahwa, A., Feng, X., Lubkeman, D., 2002. Performance evaluation of electric
distribution utilities based on data envelopment analysis. IEEE Transactions
on Power Systems 17 (3), 400–405.
Planning Commission, 2002. Annual Report (2001–02) on the Working of State
Electricity Boards and Electricity Departments. Power and Energy Division,
Planning Commission, Government of India, May Website: /http://planning
commission.nic.in/S.
PFCL, 2006. State Power Sector Performance Rating, 2004–05. Power Finance
Corporation Limited, Ministry of Power, India.
PFCL, 2008. Report on the Performance of the State Power Utilities for the Years
2004–05 to 2006–07. Power Finance Corporation Limited, Government of
India.
Ramanathan, R., 1966. An Introduction to Data Envelopment Analysis. A Tool for
Performance Measurement. Sage Publications, New Delhi, India.
LP Solver LINDO 6.1, May 2003. /http://www.lindo.comS.
Salvanes, K.G., Tjotta, S., 1990. Cost Differences in Electricity Distribution:
Economics of Scale and Economies of Density in Norwegian Electricity
Distribution Industry. WP Center for Applied Research, Oslo.
Salvanes, K.G., Tjotta, S., 1994. Productivity differences in multiple output
industries: an empirical application to power distributors. Journals of Produc-
tivity Analysis 5 (1), 23–43.
Singh, A., 2006. Power sector reform in India: current issues and prospects. Energy
Policy 34, 2480–2490.
Singh, A., 2010. Towards a competitive market for electricity and consumer choice
in the Indian power sector. Energy Policy 38, 4196–4208.
Sharma, D.P., Nair, P.S., Chandramohanan, Balasubramanian, 2005. Performance of
Indian power sector during a decade under restructuring: a critique. Energy
Policy 33, 563–576.
Smith, P., Mayston, D., 1987. Measuring efficiency in the public sector. Omega,
International Journal of Management Science 20, 569–586.
Tankha, S., Misal, A.B., Fuller, B.W., 2010. Getting reforms done in inhospitable
institutional environments: untying a Gordian Knot in India’s power distribu-
tion sector. Energy Policy 38, 7121–7129.
Thakur, T., Deshmukh, S.G., Kaushik, S.C., 2006. Efficiency evaluation of the state
owned electric utilities in India. Energy Policy 34, 2788–2804.
Thakur, T., Deshmukh, S.G., Kaushik, S.C., 2010. Performance Evaluation of Electric
Supply Utilities in India. Indiacore Publications, New Delhi, India.
Tone, K., 2002. A slacks-based measure of supper efficiency in data envelopment
analysis. European Journal of Operational Research 143, 32–41.
Totare, N.P., Pandit, S., 2010. Power sector reform in Maharashtra, India. Energy
Policy 38, 7082–7092.
Torgersen, A.M., Forsund, F.R., Kittelsen, S.A.C., 1996. Slack-adjusted efficiency
measures and ranking of efficient unit. The Journal of Productivity Analysis 7,
379–398.
UPCL, 2007–2008. State Headquarter Urjabhawan (Dehradun) and Zonal head
quarters Garhwal and Kumaon Zones, Uttarakhand Power Corporation
Limited.
Yadav, V.K., Padhy, N.P., Gupta, H.O., 2010. A micro level study of an Indian electric
utility for efficiency enhancement. Energy 35, 4053–4063.
Yang, C., Lu, W.M., 2006. Assessing the performance and finding the benchmarks of
the electricity distribution districts in Taiwan Power Company. IEEE Transac-
tions on Power Systems 21 (2), 853–861.
Zhu, J., 2000. Multi-factor performance measure model with an application to
Fortune 500 companies. European Journal of Operational Research 123,
105–124.