Utilities Policy 36 (2015) 71e78

Contents lists available at ScienceDirect

Utilities Policy
journal homepage: www.elsevier.com/locate/jup

Drivers of non-revenue water: A cross-national analysis*

Caroline van den Berg
World Bank, 1818 H Street NW, Washington DC, United States

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

Article history: Reducing water losses is seen as key to sustainable water management, but turns out to be challenging.
Received 17 November 2014 This paper applies a panel data analysis with fixed effects to assess the major drivers of non-revenue
Received in revised form water, some of which are under the control of utilities and some of which are not. The analysis finds
29 July 2015
that the most important drivers are population density per kilometer of network and type of distribution
Accepted 29 July 2015
network, which are mostly resulting from urbanization patterns which are factors mostly out of the
Available online 27 September 2015
control of the utility. Yet, low opportunity costs of water losses and high repair costs of water losses have
an important adverse effect on water loss reduction. We also found that the country environment in
JEL Classification:
H4
which the utility operates has an important impact on non-revenue water levels.
L95 © 2015 Elsevier Ltd. All rights reserved.
O57
Q25

Keywords:
Non-revenue water
Utilities
Water supply
Benchmarking
Developing countries

1. Introduction
Rapid population growth, income growth, and urbanization in
combination with a fixed supply of total renewable water resources
accelerate pressure on available per-capita renewable water re-
sources, and increase the gap between supply and demand. The
fourth World Water Development Report (2012) estimates that by
2025, 1.8 billion people will be living in countries or regions with
absolute water scarcity, and two-thirds of the world's population
could be living under water stressed conditions. Climate change is
likely to make the challenge even more daunting, as it will increase
the variability of fresh water supplies. At the same time, climate
change will increase the likelihood and frequency of water-related
disasters. The IPCC (2014) estimates that, as surface temperatures
are projected to rise, this is very likely to result in more longer-
lasting heat waves while extreme precipitation events will
become more intense and frequent in many regions.
*
The opinions reflected in this paper are the opinions of the author and should
not be attributed to the World Bank, its Executive Board of Directors, or any of its
member countries. The author would like to thank Ce
line

Nauges and Abel Mejía
for helpful suggestions.
E-mail address: cvandenberg@worldbank.org.
The water sector will have to improve the way it uses its avail-
able water resources significantly in order to deal with the chal-
lenges ahead. In the municipal sector, water productivity is less
than optimal as the difference between water put into the distri-
bution system and the amount of water billed to consumers (i.e.,
“non-revenue water”) tends to be large. For too many systems, high
levels of non-revenue water (NRW) reflect huge volumes of water
being lost through leaks (real losses) and drinking water not being
invoiced to customers (apparent losses) and unbilled authorized
consumption (Lambert et al., 2014). Kingdom et al. (2006) estimate
that the total cost to water utilities caused by NRW worldwide is
$141 billion per year. McKinsey (2013) includes water leakage
management as an important instrument for reducing the gap
between supply of and demand for municipal water. NRW reduc-
tion programs are a standard recommendation in virtually every
policy paper or report focused on improving the sustainable use of
water resources (for instance, European Community, 2006;
European Environment Agency, 2009; UNWater, 2012; and
McKinsey, 2010, 2013). The cited benefits associated with a
reduction in NRW are manifold. A reduction in water losses will
help to postpone or avoid supply investments and/or improve the
financial health of the utility, allowing it to invest in service quality
improvements. In utilities where water coverage is not universal

http://dx.doi.org/10.1016/j.jup.2015.07.005
0957-1787/© 2015 Elsevier Ltd. All rights reserved.
72 C. van den Berg / Utilities Policy 36 (2015) 71e78

and/or water is provided intermittently as is the case in many
utilities around the world, recovered water could be sold to
enhance revenues. Frauendorfer and Liemberger (2010) note that
utility owners need to be made sufficiently aware that they are
“sitting on a goldmine”; and that utility owners will need to
incentivize their staff by informing them about the level, causes,
and cost of NRW, along with the potential for improvement so that
comprehensive NRW management can be supported by the entire
organization.
If the reasons for reducing NRW are so compelling and so much
money can be saved and significant volumes of water can be used
for more productive purposes, then why has it turned out to be so
difficult to reduce NRW? The present paper looks into under-
standing the reasons why reducing NRW is so difficult.
This paper is organized as follows. In Section 2 we undertake a
literature review. In Section 3 we define NRW, review how it is
measured, and we look at the trends in NRW using the database of
the International Benchmarking Network for Water and Sanitation
Utilities (IBNET).1 This database covers financial and operational
performance data from utilities all over the world. In the fourth
section, we define a panel data model with fixed effects to analyze
what drives NRW. The estimation results of the model are pre-
sented in the fifth section, and we conclude in the last section.
In a study by Gonzalez-Gomez et al. (2011) on why NRW is so
high in developing countries, the authors conclude that the lack of
incentives is the main reason for differences in water losses be-
tween utilities. Water utilities normally lack the resources to pay
for the cost associated with a reduction of water losses, while
corruption and a lack of knowledge about NRW among users and
taxpayers further add to the lack of interest to plug water leaks. In a
follow-up study for Spain, using data from 133 municipalities in the
Andalucia region, Gonz
alez-Go
mez et al. (2012) conclude that NRW
measured as percentage of system input volumes is driven by
various physical, management, and other characteristics. They find
that the type of distribution system (gravity-fed or pumped) is a
particularly important driver of NRW for these 133 Spanish mu-
nicipalities. Saez-Fernandez et al. (2011) focus on the low oppor-
tunity cost of water losses which does not provide much incentive
for utility managers to reduce NRW. The paradox of high NRW
levels coupled with the absence of respective measures to reduce
water losses in areas that are dealing with water scarcity has also
been studied in small water distribution systems in four Mediter-
ranean countries (Kanakoudis et al., 2013b). Kanakoudis et al., 2015
also mention that tariff structure characteristics provide few in-
centives in Kazoni, Greece to reduce water losses.
3. Overview of non-revenue water

2. Literature review
3.1. Definition of NRW

Many articles have been published on NRW reduction, but they

Non-revenue water is the difference between the volume of
tend to be mostly practical guidelines as to how to reduce NRW
water put into a water distribution system and the volume that is
(Farley and Trow, 2003; and IWA Water Loss Task Force, 2003; and
billed to customers (IWA Water Loss Task Force, 2003 and
Frauendorfer and Liemberger (2010)) focusing on the processes
Kankoudis et al., 2013b). NRW is comprised of three components:
required to design and implement a NRW loss strategy. These ar-
ticles often focus on reducing NRW in developed countries. We

Physical losses include leakage from all parts of the distribution
found success stories of reducing water losses in Phnom Penh,
system and overflows at the utility's storage tanks. They can be
Cambodia (Biswas and Tortajada, 2010) and Singapore (Luan, 2010).
caused by poor operations and maintenance, the lack of active
There is also some literature on how to determine the economically
leakage control, and poor quality of underground assets;
optimal level of water losses reduction (Wyatt, 2010).

Commercial losses include customer meter under-registration,
As for what drives NRW, the literature is still rather scant, even
data-handling errors, and theft of water in various forms;
for developed countries, and dominated by an engineering

Unbilled authorized consumption includes water used by the
perspective. This is an interesting finding, as one would assume
utility for operational purposes, water used for firefighting, and
more interest in knowing more accurately what drives NRW will
water provided for free to certain consumer groups.
help in identifying the best way forward in reducing water losses.
Overall, the existing studies provide different explanations as to
what impacts NRW. Skipworth et al. (1999) show in a study of
3.2. Measurement of NRW
England, that a large number of mostly technical and environ-
mental factors affect municipal leakage rates. These factors include
The measurement of NRW is complicated. Many different in-
the age of the systems, the length and type of networks, pressure in
dicators are used to measure NRW and virtually all of them have
the systems, climate, soil conditions, traffic loading, and density of
limitations and drawbacks. The most commonly used indicator is
connections. The authors also mention that topography can explain
NRW defined as a percentage of water produced, although many
regional differences in water losses between utilities. Farley and
authors have reservations about its use (Liemberger, 2002;
Liemberger (2005) note the importance of physical factors, but
Kanakoudis et al., 2013a and Lambert et al., 2014). The Interna-
they also place considerable emphasis on management factors.
tional Water Association (IWA) generally recommends alternative
Poor management practices, poor materials and infrastructure, and
indicators, such as water losses per connection, water losses per
local social, cultural, political, and financial factors are identified as
main length, or the infrastructure leakage index (Alegre, 2006; and
NRW drivers. Their study does not delve into the details of how
Winarni, 2009). The infrastructure leakage index is a complex in-
these factors drive NRW and how this affects the design and
dicator that includes data on system pressure, length of mains,
implementation of water loss strategies. Yet, the range of factors
number and length of service connections (i.e., distance between
that drive NRW can easily be seen as factors that are either under
property line and water meter) and a water balance. This infor-
the utility's control such as management practices, while others are
mation is not necessarily easy to collect. Collecting pressure data,
less so (including physical, social, cultural and political factors).
for instance, is complicated as pressure can vary widely within a
piped water network. Lenzi et al. (2014) note that the strong
1
dependence of the ILI on average operating pressure makes it
IBNET has been developed by the World Bank with the objective to provide
comparative international benchmark performance information that can inform
sensitive to the way this parameter is defined and potentially un-
utilities and policymakers on how to improve service delivery (for more informa- suitable for comparing different systems. Recently, even more
tion, see www.ib-net.org). specific water loss indicators related to amongst others age,
 C. van den Berg / Utilities Policy 36 (2015) 71e78 73

diameter and pipe materials variables have been proposed Table 2

(Kankoudis et al., 2013a, 2013b). Yet, obtaining such information is Non-revenue water (in cubic meters per km per day).

cost-intensive, and often not available. Hence, in this paper we will Indicator 2006 2007 2008 2009 2010 2011
focus on the following accepted indicators: water losses per main Median non-revenue water 25 21 19 22 17 18
length (measured per kilometer of network) and water losses per Average non-revenue water 39 32 31 34 28 29
connection. Standard deviation 40 34 34 38 33 33
The empirical analysis is conducted using a unique database of Number of utilities reporting 1177 1440 1347 1484 1721 1605
Highest performing quartile 11 8 8 8 7 7
water utilities, namely the International Benchmarking Network
Lowest performing quartile 53 44 43 46 38 38
for Water and Sanitation Utilities (IBNET). The use of data from
Note: The 2011 data collection cycle is not yet fully complete.
IBNET ensures that the data are relatively homogeneous and
Source: IBNET database
comparable across countries. For each country, we have annual data
on utility performance. Appendix 1 provides details on the IBNET
database. As shown in Table 3, water losses vary significantly between
For the sample, the median NRW (as measured by the volume of utilities as is expected. In better performing utilities, median water
water lost as a percentage of water produced) was 32 percent in production stands at 207 L per capita per day (lcd) compared to
2006 and 29 percent five years later in 2011 (Table 1), while the 291 lcd in utilities with higher water losses. In utilities with low
standard deviation remains high. water losses, consumption levels are 173 lcd compared to 180 lcd
The median NRW (as measured by water losses per kilometer for utilities with higher water losses. Table 4 shows that the lower
per day, abbreviated to NRWkmday) decreased from 25 cubic me- the water production, the lower NRW (as measured by
ters per km in 2006 to 18 in 2011. Yet, this indicator shows wide NRWkmday).
variations by year and among utilities (Table 2). In 2011, the best We look also into a set of indicators that may reflect the man-
performing quartile of utilities had a median NRWkmday of 7 cubic agement quality of the utility. The quality of financial management,
meter or less. The worst performing quartile of utilities were those as measured by the operating cost coverage ratio,2 is 1.26 for util-
where the median value was 38 or higher. ities with low water losses compared to 1.23 for utilities with high
The different NRW indicators are not necessarily strongly water losses. The difference is very small, but statistically signifi-
correlated. A utility may show excellent performance using one cant at the 10 percent level, and shows that utilities with lower
NRW indicator but poorer performance when measured by water losses are able to cover a larger part of their operation and
another. An example is Singapore's Public Utilities Board which maintenance costs with their operating revenues. We also calcu-
with a NRW of only 4 percent of system input in 2007 can be lated a variable that showed the absolute buffer that utilities have
considered an excellent performing utility. When using NRW per by calculating the net operating revenues per cubic meter of water
km per day, however, Singapore's performance falls below the best sold. Utilities with low water losses are able to generate net reve-
performing quartile of utilities. Similarly, some utilities show high nue of USD0.35 per cubic meter of water sold compared to only USD
NRW when measured as a percentage of water produced, but the 0.14 in utilities that cope with higher water losses. It is not clear,
losses look much less dramatic when using another indicator. however, whether the larger financial buffer results from lower
Hence, the IWA recommends looking at more than one loss indi- water losses or whether better financial performance enables
cator when evaluating utility performance. In other words, focusing utilities to deal more effectively with water losses.
on only one indicator may be misleading (Liemberger, 2002). It is often assumed that metering is a prerequisite for reducing
In Table 3, we present descriptive statistics for utilities with low water losses. The database allows us only to report on customer
NRWkmday (defined as those utilities in the best quartile of the metering as data on production metering is not available. Table 3
sample) and those that register higher levels of NRW. Mean com- shows that utilities with higher water losses tend to have lower
parison tests show that low levels of water losses tend to occur in levels of customer metering (80 percent) than those with lower
utilities that are smaller in terms of people served, kilometers of water losses (87 percent). The difference is statistically significant,
network, number of towns served, and have a lower population but the difference in levels of metering is rather small. The effect of
density (i.e., people served per kilometer of network). The average universal metering tends to be much more significant. About 57
network length in utilities with lower water losses is 737 km, percent of the utilities with low water losses tend to be fully
compared to 1037 km in utilities that are characterized by higher metered, compared to 41 percent of the utilities with high water
water losses. In the case of utilities with lower water losses, 81 losses. In many developing countries that dominate our sample,
percent of the better performing utilities serve only one town utilities do not always supply water on a 24/7 basis. The number of
compared, to 72 percent for the utilities with higher water losses. hours utilities are able to supply water is slightly higher (and sta-
Utilities with low water losses are also faced with lower population tistically significant) in utilities with low water losses. Gonza
lez-
densities (i.e., the number of people served per kilometer of water Go
mez et al. (2012) considered whether private-sector participa-
supply network). The average number of people served per kilo- tion is a factor in explaining water losses. We do not find a statis-
meter of network is 179 in utilities with low water losses compared tically significant difference here. Utilities with higher forms of
to 450 in utilities with high water losses. regulatory oversight tend to show lower water losses. Yet, in case of
staff productivity (as measured by the number of staff per 1000
people served), utility with lower water losses have 1.74 staff per
Table 1 1000 people compared to 1,34 for utilities with higher water losses.
Non-revenue water (as percentage of water production).

Indicator 2006 2007 2008 2009 2010 2011

Median non-revenue water 33 30 29 29 30 29 2

The operating cost coverage ratio measures in how far total operation and
Average non-revenue water 35 32 31 31 31 31
maintenance costs are covered by operating revenues. If the value is larger than
Standard deviation 18 17 17 18 18 17
one, the utility is able to generate sufficient revenues to cover its basic operation
Number of utilities reporting 1180 1452 1365 1383 1725 1589
and maintenance costs. The larger the value is the more able the utility is to cover
Note: The 2011 data collection cycle is not yet fully complete. not only operation and maintenance costs but also increasingly the cost of depre-
Source: IBNET database ciation and debt service).
74 C. van den Berg / Utilities Policy 36 (2015) 71e78

Table 3
Non-revenue wateredescriptive statistics for NRW.

Variable Utilities with low NRW Utilities with high t-value

losses NRW losses

NRW in percent 17.0 36.4 51.23*

NRW in km per day (in cum) 4.0 41.3 47.98*
NRW per connection per day (in cum) 0.16 0.88 7.78*
Physical factors
Connection density (number of connections per km of network) 48.94 84.89 15.15*
Population per Connection 6.23 11.36 3.57*
Population Density (Population served per km of network) 179 450 9.45*
Length of network 737 1037 3.67*
Full water coverage (dummy) 0.29 0.24 4.10*
Number of towns served 14 12 1.13
Utilities only serving one town (dummy) 0.81 0.72 7.64*
Service area is urban (dummy) 0.47 0.66 14.25*
Size of utility in people served (1 ¼ smallest, 6 ¼ largest) 2.23 3.12 25.20*
Share of energy costs in total operation and maintenance costs 0.27 0.26 1.52***
Water production per capita per day (lcd) 207 291 24.57*
Water consumption per capita per day (lcd) 173 180 2.53*
Management and institutional factors
Operating cost coverage ratio 1.26 1.23 1.60***
Net revenues (USD) per cubic meter sold 1.26 1.23 1.74**
Presence of PSP (dummy) 0.21 0.20 0.93
Universal metering (dummy) 0.57 0.41 11.16*
Number of staff per 1000 people served 1.74 1.34 7.56*
Hours of supply 22.54 22.13 3.17*
Energy cost per cum of water produced 0.15 0.09 27.84*
Planning process (1 ¼ annual budgets; 2 ¼ multi-annual budgets, 3 ¼ other 1.54 1.58 1.15
Share of industry in total volume sold 0.35 0.31 4.80*
Oversight on service and price (1 ¼ government department; 2 ¼ independent board of stakeholders; 1.63 1.42 5.26*
3 ¼ independent regulator; 4 ¼ other)
Other factors
Wage rate levels in utility (USD) 10,063 9306 3.36*
GDP per capitaa 12,568 6633 30.38*
Level of Corruptionb (1 ¼ highest, 10 ¼ lowest) 4.33 3.47 32.16*

* Significant at 1 percent level.

** Significant at 5 percent level.
*** Significant at 10 percent level.
a
GDP per capita and level of corruption is based on country level data.
b
Level of corruption: the higher the value of the variable, the lower corruption (1 ¼ highest level of corruption, 10 ¼ lowest level of corruption).

Table 4
NRW by water production category.

Water production category Median water production Median NRW in percent Median NRW per km per day

<200 lcd 146 25.8 10.3

200e400 lcd 276 31.8 23.9
400e600 lcd 465 35.7 49.4
>600 lcd 685 36.4 70.3

Source: IBNET database

Finally, the GDP per capita in utilities with lower water losses is
almost double that of GDP per capita in utilities with high water
losses. Table 3 shows that utilities with lower water losses are more
likely to be located in countries where corruption levels also are
lower. Corruption is likely to have a significant effect on water
losses as it tends to negatively impact infrastructure investment as
well as maintenance, procurement processes, and billing practices
as outlined by Transparency International (2008). As seen in
Table 5, water losses show a particular pattern in which they are
high in low-income countries, higher still in lower-middle income
countries, then lower once countries reach upper middle-income
status, and lower still when countries gain high-income status.
Variation within the four country categories tend to be large, and
hence local factors play an important role in explaining NRW levels.
4. Specification of NRW loss model
Following the still limited literature on the drivers of water
losses, we will specify the water loss function as
L ¼ f (P, M, O) þ e þ u (1)
This formula describes the relationship between non-revenue
water as measured by water losses (L) on the one hand, and a
vector of physical characteristics of the water infrastructure (P), a
vector of management characteristics of the utility (M), and a vector
of other factors (O) that characterize the area in which the utility is
located. As we depend on panel data that covers the period be-
tween 2006 and 2011 and includes utilities from a subset of
countries, we include the country effects (e) and an error term (u).
Each country has its own characteristics that may influence the
water losses. These country effects may refer to among others
institutional, legal, and regulatory frameworks; fiscal policies;
environmental, public health, engineering and building standards;
and levels of corruption.
The estimation of the single loss reduction function provides
 C. van den Berg / Utilities Policy 36 (2015) 71e78
Table 5
NRW as measured in cubic meters per kilometer per day by Region.
Indicator 10th Percentile 25th Percentile
Low-income Countries 5.35 9,95
Lower-middle income countries 5.14 11.94
Upper-middle income countries 3.16 8.11
High income countries 1.86 3.29
Source: IBNET database.
insight on variables that might drive water losses. We will specify
water losses and the explanatory variables in natural logarithm
terms, so that the coefficients will measure the elasticity of water
losses, i.e., by how much losses will change relative to a change in
the different drivers.
In the model we will use NRWkmday as the dependent variable.
The reason for selecting this specific indicator of water loss is
mainly that it is more consistently defined than NRW per connec-
tion. What constitutes a connection can be differently defined by
country. In Latin America, for instance, the number of people per
connection tends to be low, reflecting household size. In other parts
of the world, including some African countries, households share
connections with neighbors. In Eastern Europe and Central Asia,
multi-family buildings are often fitted with a single water
connection and hence the number of people per connection can be
very large.
Our first hypothesis is that physical characteristics of water
systems affect the level of water losses. The factors included in the
model are population density (per kilometer of network as
measured by LN_POPDENS). We assume that the higher the pop-
ulation density in the network, the higher the water losses will be.
A denser network tends to be linked with more pressure re-
quirements to reach the higher floors of buildings. Pressure is e
with age and materialseamong the most important factors influ-
encing the level of leakage in distribution systems (Skipworth et al.,
1999). Population density is a factor that is not directly under the
control of the utility. Prieto et al. (2010) show that scale and density
economies are important in the provision of water and wastewater
services especially those related to urban dispersion. Their results
show the existence of latent economies of scale and density due to
the effect of urban sprawl resulting in higher costs in the provision
of infrastructure. Hence, we include a variable that measures the
urban nature of the network (URBANDUMMY) which we assume to
be positively correlated with water losses. The share of energy costs
as part of total operation and maintenance costs (LN_ENERGY-
SHARE) is used as a proxy for the type of the distribution system. It
is assumed that if a system requires more pressure and hence
pumping, the higher the share of energy costs will be. We assume
that more pumping translates into higher water losses consistent
with the definitions of the Infrastructure Leakage Index
(Liemberger, 2010; Lambert et al., 2014).
The second hypothesis is that utility management quality affects
water losses. It is assumed that the better the quality of utility
management, the lower water losses. We assume that universal
water coverage provided 24 h day (FULLSERVICE) is a factor under
management control. Water systems tend to be designed for a
horizon of several decades and less mature water systems may have
significant excess production capacity (depending on a country's
standards and regulations). As access to water supply services im-
proves (in terms of people and consumption), excess capacity tends
to decline. Thus, water losses should be relatively lower in water
networks that serve the total population 24 h a day (Table 4).
We also include a measure of financial management of the
utility defined as NETREV_USD_CUM, which measures the differ-
ence between operating revenues and operation and maintenance
75
50th Percentile 75th Percentile 90th Percentile
18.22 34.28 66.70
27.28 52.35 91.62
21.18 44.55 76.46
7.04 14.26 24.88
costs per cubic meter of water sold. This variable can take on a
positive value if the utility is able to cover its basic operation and
maintenance costs through its operating revenues, but can also
take on a negative value. We assume that the more financial buffer
a utility has, the higher the likelihood of lower water losses as the
utility has more financial capacity to deal with water losses.
A third management variable is the number of staff per 1000
people served (LN_STAFFPEOPLE), which is an indication of how
efficient a utility is in managing its business, including managing its
water losses. The more staff per 1000 people served, and hence the
lower the staff productivity, the higher the water losses are ex-
pected to be.
Finally, we include a variable that according to Garcia and
Thomas (2001) reflects the opportunity cost of water losses as
measured by the marginal cost of increasing production associated
with higher energy3 and resource costs (in most systems, the
resource costs of drinking water are negligible). The authors point
out that reductions in NRW are constrained by high repair costs,
noting that most utilities have two options when they need to
produce and distribute more water to meet demand. The first op-
tion is to keep production constant while reducing NRW through
more frequent repair and maintenance; the second option is to
increase production. When the labor and capital costs for repair and
maintenance are higher than the energy and resource costs asso-
ciated with increased production (i.e., the opportunity costs of
water losses), the second option will be the preferred one. We as-
sume that the higher the opportunity cost of water losses
(LN_WATERLOSS), the lower water losses will be. This variable is
somewhat similar to the variable used by Saez-Fernandez et al.
(2011).
The last hypothesis states that also other factors affect the level
of water losses. Data on GNI and corruption are only available at the
national level and thus captured in the fixed effects of the model.
We include the wage rate per employee in USD as a proxy for in-
come levels at the utility level as reflected in LN_WAGERATE. We
expect that the higher the wage rate will be, and hence the higher
the cost of repair and maintenance, the higher the likelihood of
higher water losses.
5. Estimation results
The analysis was conducted using cross-sectional data from
utilities covering 63 countries. This sample is smaller than the
available sample, as we cleaned the data set for outliers based on 1
percent at either tail of the distribution. Given the nature of the
dependent variable, we used a panel data regression model with
fixed effects, with water losses measured by LN_NRWKM and
controlling for heteroskedasticity. Table 6 shows the results of the
panel-data regression with fixed effects. Estimated coefficients are
all significant and have the expected sign as can be seen in Table 6,
3
The variable costs should theoretically also include chemical costs, but chemical
cost data are not included in the dataset. In most utilities, labor and energy make up
the bulk of the operation and maintenance costs.
76 C. van den Berg / Utilities Policy 36 (2015) 71e78

It also shows that the fixed effects variable captured by the country definitions of economic development of low, lower-middle, upper-
effects is very significant. The overall fit of the model is satisfactory, middle and high-income countries, with results shown in Table 7.
with a R square adjusted of 0.49. The errors in the model are not There are notable differences between the four models; the co-
correlated with the regressors in the fixed-effects model. Almost 50 efficients differ at time significantly, yet all of the signs for the co-
percent of the variance in the model is due to differences across efficients are similar in direction. Population density is more or less
panels (or countries), showing that the characteristics of the similar among the different income groups, and one of the most
country have a significant effect on water losses. Hence the envi- important drivers of NRW.
ronment in which the utility operates appears to play a key role in
explaining water losses. 6. Concluding remarks and policy recommendations
We find that the higher the population density of the network,
the higher the water losses. The effect of this variable is very sig- The literature identifies large benefits from a reduction of water
nificant, with a t-value of greater than 37. The coefficient is also losses (Kingdom et al., 2006; Frauendorfer and Liemberger, 2010;
large showing that a 10 percent increase in connection density is McKinsey, 2013). Nevertheless, in the past six years, based on
associated with an increase in losses of 10.2 percent. URBAN- various measures, improvements in reducing water losses have
DUMMY is also positively related with water losses. A utility with been limited.
an urban service area appears to more likely result in higher water The main empirical finding of this paper is that the key drivers of
losses. An increase in LN_ENERGYSHARE also appears to be asso- water losses (using a large sample of utilities located in 63 coun-
ciated with higher water losses. The effect is very significant and a tries) are partly linked to the physical characteristics of the water
10 percent increase in the energy share, is associated with a 6.6 supply system. Population density and the nature of the service
percent increase in water losses. area, which mainly result from urbanization and land settlement
Turning to the role of management factors in explaining water patterns, are important drivers of NRW. Also, the nature of the
losses, FULLSERVICE has a negative impact on water losses. The distribution network (pumped or gravity as proxied by the share of
higher the quality of service (i.e., water supplied to all of the pop- energy costs in total operation and maintenance costs) is mostly
ulation in the service area 24 h per day), the lower water losses will determined by geography. It appears that utilities located in highly
be. NETREV_USD_CUM which measures the difference between urbanized and dense settlements (especially those where water
total revenues and total operation and maintenance costs per cubic distribution depends on pumping water) are likely to experience
meter of water sold shows a negative relationship with water los- higher levels of NRW. Reducing non-revenue water in such envi-
ses. The higher the net revenues per cubic meter of water sold, and ronments may not be easy, especially if the process of urbanization
hence the more financial buffer, the lower water losses will be. is ongoing. Differences in the context in which the utility is oper-
LN_STAFFPEOPLE shows a positive relationship with water losses. ating makes a “one size fits all” benchmark for what constitutes an
An increase in the number of staff per 1000 people served by 10 acceptable level of NRW not very useful, as the cost of reducing
percent will increase water losses by 4.4 percent. Finally, the op- water losses can differ significantly between utilities. A fact
portunity cost of water losses as measured by the energy compo- implicitly acknowledged in the ILI index that calculates different
nent of costs (as measured by LN_WATERLOSS) are negatively NRW benchmarks for different pressure levels.
related to water losses. A 10 percent increase in the opportunity Some factors that drive water losses are within the control of
cost of water losses results is associated with water losses that are utilities. Water utilities have room to manoeuver water losses by
almost 6.7 percent lower. improving how they operate their systems. Utilities that at least
LN_WAGERATE shows that a higher wage rate at the utility level cover their operation and maintenance expenses through revenues
is associated with higher water losses. An increase in staff wage face lower levels of NRW.
rates by 10 percent will increase water losses by 5.7 percent. This Yet even more important are the opportunity cost of water
might be expected as higher wage rates may be linked to higher losses, and the staff productivity (and its effect on the cost of repair
repair and maintenance costs associated with skilled labor needed and maintenance of water losses).
for water loss reduction. A final set of conclusions concern the sometimes paradoxical
We ran the model again for utilities located in countries with nature of improving water utility performance. Water loss reduc-
different levels of development. We use the World Bank defined tion is seen as a key ingredient to improving utility performance,
but the analysis shows that often proposed utility performance
tools and water loss-reduction do not always go hand in hand. For
instance, energy efficiency programs can improve the utility's
bottom line, but an energy efficiency improvement will also reduce
Table 6
the opportunity cost of water losses. Larger utilities also have larger
Results of the panel data fixed effects regression.
water losses, a finding that seems to deflect from scale economies.
Variable LN_NRW_KM Coefficient Standard error T-value P-value However, this finding is consistent with research by Nauges and
LN_POPDENS 1.017 0.027 37.65 0.000 van den Berg (2008) and Gonz

mez et al. (2012) showing
alez-Go
URBANDUMMY 0.139 0.049 2.83 0.006 that economies of scale are more evident in smaller companies and
LN_ENERGY_SHARE 0.665 0.108 6.17 0.000 tend to diminish with large operations.
LN_WATERLOSS 0.673 0.083 8.12 0.000
FULLSERVICE 0.084 0.027 3.11 0.003
Policy Recommendations. NRW is impacted by factors that
NETREV_USD_CUM 0.215 0.067 3.22 0.002 utilities can control and factors beyond the control of utilities (such
LN_STAFFPEOPLE 0.435 0.078 5.57 0.000 as settlement patterns in urban areas and topography). Under-
LN_WAGE_RATE 0.566 0.100 5.68 0.000 standing what drives NRW is important in order to design proper
Constant 8.653 0.937 9.24 0.000
water loss intervention. Key factors in setting these incentives
Number of observations 4515
Number of groups 63 appear to be the opportunity cost of water losses (where a value is
R square 0.4943 placed on wasting water), staff productivity and assurances that
F test 565.16 utilities can operate and maintain their infrastructure. It is also
Prob > F 0.0000 important that as many people still do not have access to piped
Rho 0.486
water, utilities properly plan for water supply infrastructure and
 C. van den Berg / Utilities Policy 36 (2015) 71e78 77

Table 7
Results of the panel data fixed effects regression for utilities by income group.

LN_NRW_KM Low income Lower-middle income Upper-middle income High income

Coeff p- value Coeff p- value Coeff p- value Coeff p- value

LN POPDENS 1.037 0.000 1.019 0.000 1.029 0.000 0.961 0.000

URBANDUMMY 0.199 0.000 0.196 0.001 0.060 0.148 0.830 0.000
LN_ENERGYSHARE 0.693 0.000 0.811 0.000 0.607 0.000 0.701 0.000
LN_WATERLOSS 0.737 0.000 0.801 0.000 0.628 0.000 0.522 0.000
FULLSERVICE 0.285 0.047 0.169 0.010 0.090 0.013 0.086 0.531
NETREV_USD_CUM 0.658 0.000 0.457 0.000 0.170 0.000 0.256 0.013
LN_STAFFPEOPLE 0.647 0.000 0.579 0.000 0.331 0.000 0.675 0.000
LN_WAGE_RATE 0.602 0.000 0.556 0.000 0.559 0.000 0.350 0.001
Constant 9.780 0.000 8.698 0.000 8.577 0.000 5.896 0.000
Number of observations 477 1069 2821 143
Number of groups 21 26 24 4
R square 0.6233 0.4715 0.5102 0.7584
F test 124.12 171.32 266.06 28.00
Prob > F 0.000 0.000 0.000 0.000
Rho 0.532 0.537 0.260 0.311

consider how design standards and settlement patterns can be
taken into consideration when planning new infrastructure.
Thirdly, when setting benchmarks for NRW, it is important to
provide ranges of good performance instead of absolute bench-
marks, as external factors outside the control of the utility can make
it more or less difficult to achieve success. Finally, the country
environment in which a utility operates appears to influence
effectiveness in controlling water losses. Country effects may relate
to labor and energy policies, urbanization and land settlement, and
engineering design standards and norms, among others. Follow-up
work is needed to understand which of these factors contribute
significantly to lower water losses.
Future Directions of Research. In view of the unobserved
characteristics that we found, future research may focus on getting
more insight into variables that might be important in explaining
water losses, but for which we did not have data, such as topog-
raphy, age of system, soil conditions, and the production capacity of
the water supply systems. In addition, more data on energy vol-
umes and prices and abstraction fees would be useful to gain
insight on the opportunity cost of water losses. Utilities focusing on
meeting new demand have an incentive to increase water pro-
duction when the costs of repair exceed the cost of production. It is
much easier to increase production in systems with overcapacity.
Hence, further research is warranted in how engineering design
standards and norms affect water losses and loss reduction. Finally,
as the country effects are significant, more research is needed in
terms of unbundling these effects, as they may include a wide va-
riety of factors such as sector regulation, labor legislation and
policies, environmental regulation, design standards, energy pol-
icies, and others.
Appendix. Appendix on IBNET Database
We used a unique database that collects data on the financial
and operational performance of water utilities, namely the Inter-
national Benchmarking Network for Water and Sanitation Utilities
(IBNET) at www.ib-net.org. IBNET ensures that the data used in this
analysis are relatively homogeneous and comparable across coun-
tries. For each country, we have annual data on different aspects of
the utilities’ operational and financial performance. The IBNET
database contains information on performance from 1982 water
utilities in 2010 serving 583 million people with water in 18,308
cities and towns. This is approximately 16 percent of the total
population of all households with piped water access in the world.
The database represents the equivalent of more than US$46 billion
in annual revenue in 2010. The database includes municipal water
and/or wastewater utilities, regional and national utilities.
Participation in IBNET is voluntary. This may affect the sample as
it is possible that better performing utilities are more likely to
report results than poor performers. IBNET works with national
utility associations, regulatory agencies, and ministries that collect
data on utility performance. Hence in many countries, data are
collected from a range of utilities which are likely to include better
and worse performing utilities. The voluntary character may also
result in changes in the database over time, as sometimes utilities
participating in one round of data collection do not necessarily
participate in a next round of data collection. In addition, the data-
collection process takes time as the data is collected once annual
reporting has taken place while the cost associated with data
collection does not allow for annual updates. In this analysis, we
include data from 2006 to 2011 when the number of utilities in the
database is relatively stable and the sample size does not differ
significantly by year.4
IBNET data is collected from utilities mainly through national
utility or regulatory associations using a standard data tool
(including standard definitions) to ensure that data are relatively
comparable across utilities. Even though definitions are standard-
ized, the quality of the data collection can differ significantly across
utilities. Information is collected from countries in all the world's
regions. Data from countries in Africa and South Asia, which tend to
be categorized as low income, are less-well represented for the year
2010 as data collection has not yet been completed, while data
collection is especially complicated in South Asia where utilities are
often municipal departments, and reporting in such departments is
less well-established. In the past years, economic development has
been rapid especially in emerging countries. The average per-capita
nominal Gross Domestic Product (GDP) increased rapidly from
US$4937 in 2006 to US$9567 in 2011. This rapid economic growth
has had major consequences, reducing the proportion of people
living in low-income countries as many countries have moved to
higher income categories. As a result, the income classification in
the IBNET sample has changed significantly too. For more details,
reference is made to Danilenko et al. (2014).
References
Alegre, Helen, 2006. Performance Indicators for Water Supply Services. IWA
Publishing.
Biswas, A.K., Tortajada, C., 2010. Water supply of Phnom Penh; an example of good
4
The IBNET database has been growing over time. It started with a small number
of utilities in 1996 and has been growing to 1982 utilities in 2010.
78 C. van den Berg / Utilities Policy 36 (2015) 71e78
governance. Int. J. Water Resour. Dev. 26, 157e172.
Danilenko, Alexander, van den Berg, Caroline, Macheve, Berta, Joe Moffitt, L., 2014.
The IBNET Water Supply and Sanitation Blue Book 2014. The International
Benchmarking Network for Water and Sanitation Utilities Databook, Wash-
ington, D.C.
European Community, 2006. Water Scarcity Management in the Context of Water
Framework Directive. Brussels.
European Environment Agency, 2009. Water Resources across Europe. Confronting
Water Scarcity and Drought. Technical Report 2/2009. European Environment
Agency, Copenhagen.
Farley, M., Liemberger, R., 2005. Developing a non-revenue water reduction strat-
egy: planning and implementing the strategy. Water Sci. Technol. Water Supply
5 (1), 41e50.
Farley, Malcolm, Trow, Stuart, 2003. Losses in Water Distribution Networks: a
Practitioner's Guide to Assessment, Monitoring and Control. IWA Publishing.
Frauendorfer, Rudolf, Liemberger, Ronald, 2010. The Issues and Challenges of
Reducing Non-revenue Water. Asian Development Bank.
Garcia, S., Thomas, A., 2001. The structure of municipal water supply costs: appli-
cation to a panel of French local communities. J. Prod. Analysis 16, 5e29.
Gonza
lez-Go
mez, F., García-Rubio, M.A., Guardiola, J., 2011. Why is non-revenue
water so high in so many cities? Int. J. Water Resour. Dev. 27, 345e360.
Gonza
lez-Go
mez, F., Martínez-Espin ~ eira, Roberto, García-Valin~ as, María A., García-
Rubio, Miguel A., 2012. Explanatory factors of urban water leakage rates in
southern Spain. Util. Policy 22, 22e30.
International Benchmarking Network for Water and Sanitation Utilities (IBNET) at
www.ib-net.org.
IPCC, 2014. In: Core Writing Team, Pachauri, R.K., Meyer, L.A. (Eds.), Climate Change
2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth
Assessment Report of the Intergovernmental Panel on Climate Change. IPCC,
Geneva, Switzerland.
IWA Water Loss Task Force, 2003. Assessing Non-revenue Water and its Compo-
nents: a Practical Approach. Water 21.
Kanakoudis, V., Tsitsifli, S., Samaras, P., Zouboulis, A., Banovec, P., 2013a. A new set of
water-loss related performance indicators focused on areas facing water scar-
city conditions. Desalination Water Treat. 51 (13e15), 2994e3010.
Kanakoudis, V., Tsitsifli, S., Samaras, P., Zouboulis, A., 2013b. Assessing the perfor-
mance of urban water networks across the EU mediterranean area: the paradox
of high NRW levels and absence of respective reduction measures. Water Sci.
Technol. Water Supply 939e950.
Kanakoudis, V., Tsitsifli, S., Kouziakis, C., Lappos, S., 2015. Defining the level of non-
revenue water in Kozani, Greece: is it a typical case? Desalination Water Treat.
54 (8), 2170e2180.
Kingdom, B., Liemberger, R., Marin, P., 2006. The challenge of reducing non-revenue
water (NRW) in developing countries. How the private sector can help: a look at
performance-based service contracting. In: Water Supply and Sanitation Sector
Board Discussion Paper Series, Paper 8World Bank, Washington, D.C.
Lambert, A., Charalambous, B., Fantozzi, M., Kovac, J., Rizzo, A., Galea St John, S.,
2014. 14 Years Experience of Using IWA Best Practice Water Balance and Water
Loss Performance Indicators in Europe at. www.leakssuite.com.
Lenzi, C., Bragalli, C., Bolognesi, A., Fortini, M., 2014. Infrastructure leakage index
assessment in large water systems. Procedia Eng. 70, 1017e1026.
Liemberger, Ronald, 2002. Do You Know How Misleading the Use of Wrong Per-
formance Indicators Can Be? Paper Presented at the IWA Managing Leakage
Conference, Cyprus, November 2002.
Liemberger, Ronald, 2010. Recommendations for Initial Non-revenue Assessment.
IWA Water Loss, 2010 at. http://www.miya-water.com/user_files/Data_and_
Research/miyas_experts_articles/15jun2010/Recommendations%20for%
20Initial%20Non-Revenue%20Water%20Assessment.pdf.
Luan, Ivy Ong Bee, 2010. Singapore water management policies and practices. Int. J.
Water Resour. Dev. 26 (1), 65e80.
McKinsey and 2030 Water Resources Group, 2010. Charting Our Water Future:
Economic Frameworks to Inform Decision-making.
McKinsey Global Institute and McKinsey Infrastructure Practice, 2013. Infrastruc-
ture Productivity: How to Save $1 Trillion a Year.
Nauges, Ce
line, van den Berg, Caroline, 2008. Economies of density, scale and scope
in the water supply and sewerage sector: a study of four developing and
transition economies. J. Regul. Econ. 34 (2), 144e163.
Prieto, Angel M., Zofio, Jose L., Alvarez, Immaculada, 2010. Urban Patterns, Popu-
lation Density and Optimal City Dimension: the Case of Public Infrastructure.
Economic Policy Working Paper Series 8/2010. Universidad Autonoma de
Madrid.
Saez- Fernandez, Francisco, Gonza
lez-Go
mez, Francisco, Picazo-Tadeo, Andres, 2011.
Opportunity costs of ensuring sustainability in urban water services. Water
Resour. Dev. 27 (4), 693e708.
Skipworth, P.J., Saul, A.J., Machell, J., 1999. The effect of regional factors on leakage
levels and the role of performance indicators. Water Environ. J. 13, 184e188.
UNWater, 2012. Managing Water under Uncertainty and Risk. World Water
Development Report 4 at. http://www.unwater.org/publications/publications-
detail/en/c/202715/.
Transparency International, 2008. Global Corruption Report: Corruption in the
Water Sector. Cambridge University Press.
Winarni, W., 2009. Infrastructure leakage index (ILI) as water losses indicator. Civ.
Eng. Dimens. 11, 126e134.
Wyatt, A., 2010. Non-revenue Water: Financial Model for Optimal Management in
Developing Countries. RTI Press publication No. MR-0018-1006. Research Tri-
angle Park, NC.