How well is the demand-driven, community management model
for rural water supply systems doing? Evidence from Bolivia,
Peru and Ghana
, Jennifer Davis
, Linda Prokopy
, Kristin Komives
, Heather Lukacs
, Alexander Bakalian
and Wendy Wakeman
Corresponding author. Department of Environmental Sciences & Engineering, Rosenau CB#7431, School of Public Health,
University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599, USA
Tel: 1-919-966-7645. E-mail: Dale_Whittington@unc.edu
Manchester Business School, University of Manchester, Manchester, UK
Department of Civil & Environmental Engineering & Woods Institute for the Environment, Stanford University,
Terman Engineering M-21, Stanford, CA 94305-4020, USA
Department of Forestry and Natural Resources, Purdue University, 195 Marsteller Street, West Lafayette,
Indiana 47907-2093, USA
Institute of Social Studies, Kortenaerkade 12, The Hague, 2518 AX, The Netherlands
WaterPartners International, 2405 Grand Blvd., Ste. 860, Box 12, Kansas City, MO 64108, USA
The World Bank, 1818 H Street, NW, Washington, DC 20433, USA
This paper reports the main ﬁndings of a multi-country research project designed to develop a better
understanding of the performance of community-managed rural water supply systems in developing countries.
Data were collected from households, village water committees, focus groups of village residents, system
operators and key informants in 400 rural communities in Peru, Bolivia and Ghana. Our ﬁndings suggest that the
demand-driven, community management model, coupled with access to spare parts and some technical expertise,
has come a long way toward unraveling the puzzle of how best to design and implement rural water supply
programs in developing countries. In all three countries, rural water supply projects were working. Among the
households included in our sample in Peru and Bolivia, 95% had operational taps at the time of our ﬁeld visit. In
90% of the villages in Ghana, all project handpumps were still working. Not only had the rural water systems not
broken down, but almost all the households in these communities were obtaining at least some of their water from
the systems. However, some households were also still using water from other sources. In Ghana, 38% of
households still reported using water from unprotected sources (e.g. springs, river, open wells) for drinking and/or
cooking. Another troublesome ﬁnding is that rural households in the sample villages are paying very little for the
improved water services and, as a result, the ﬁnances of many village water committees are in poor shape.
Water Policy 11 (2009) 696–718
qWorld Bank 2009
Keywords: Bolivia; Community management; Ghana; Peru; Post-construction support; Rural water supply
This paper reports the main ﬁndings of a large, multi-country research project designed to develop a
better understanding of the performance of rural water supply systems in developing countries. We
initiated this research project in 2004 to investigate how the provision of support for communities after
the construction of a rural water supply project affected its medium-term performance, deﬁned as the
period of system operation 3– 12 years after initial construction. We collected information from
households, village water committees, focus groups of village residents, system operators and key
informants in 400 rural communities in Peru, Bolivia and Ghana. In total we discussed community water
supply issues with approximately 10,000 individuals in these communities.
In the course of our investigation of the effects of post-construction support (PCS), we also learned
much about how rural water supply systems are faring in parts of Peru, Bolivia and Ghana. We believe
these observations will be of broad policy interest to professionals in the water sector because they
contradict the general perception that most rural water systems fail and that success is dependent on cost
recovery through sizeable, on-going user contributions. This paper summarizes these more general
ﬁndings about the status of rural water projects, as well as our principal conclusions about the effects of
PCS services on system performance
In the next, second section of this paper we discuss the policy context for rural water supply programs
in developing countries and summarize the conventional wisdom about how rural water supply programs
should be designed and implemented to ensure their sustainability. The third section describes our
research design and the fourth section the ﬁeldwork conducted in Peru, Bolivia and Ghana. The ﬁfth
section presents our main ﬁndings regarding the status and performance of existing rural water projects
and the factors associated with their success. In the sixth and ﬁnal section we discuss some of the
implications of our ﬁndings.
In the 1980s it became widely recognized among sector professionals that many rural water supply
programs in developing countries were performing poorly (Churchill et al., 1987;Therkildsen, 1988;
Briscoe & DeFerranti, 1988). Regardless of the type of technology utilized, systems were not being
repaired and were falling into disuse. Cost recovery was minimal and revenues were often insufﬁcient to
pay for even operation and maintenance, much less capital costs. Communities did not have a sense of
ownership in their water projects and households were not satisﬁed with the projects that donors and
national governments installed.
An intensive discussion ensued within the water resources profession about the reasons why success
in the rural water supply sector was so difﬁcult to achieve. Engineers blamed poor quality construction,
The more detailed results about the effects of post-construction technical assistance can be found in our papers and reports on
Ghana, Bolivia and Peru (Komives et al., 2007;Thorsten, 2007;Davis et al., 2008;Prokopy et al., 2008).
D. Whittington et al. / Water Policy 11 (2009) 696–718 697
anthropologists described a lack of community participation, political scientists reported rent-seeking
and poor governance structures and economists complained of poor pricing and tariff design. In the
1990s a consensus emerged that pre-project planning procedures for rural water supply programs needed
to be more “demand-driven”. The necessary components of a demand-driven process differ somewhat
depending on who one asks, but most would agree that project planning should (1) involve households in
the choice of both technology and institutional and governance arrangements, (2) give women a larger
role in decision-making than has historically been the norm and (3) require households to pay all of the
operation and maintenance costs of providing water services and at least some of the capital costs (Sara
et al., 1996;Sara & Katz, 1997;Whittington et al., 1998).
The rationale for involving households in the choice of technology was to ensure that engineering
designs were responsive to local needs and realities. Women should assume a greater role in decision-
making because they were the ones who best knew these local realities and were primary beneﬁciaries of
the projects. Cost recovery through user fees served three purposes. First, it required households to pay
for services once operational, provided revenues to keep the system running and reduced dependence on
higher-levels of government. Second, charging households (part of) the capital cost of system
construction prior to installing the water supply system established a “demand ﬁlter” that theoretically
prevented water systems from being built in communities where they were low-priority development
projects. Third, requiring capital contributions by communities was expected to foster a sense of
community ownership of the facilities, which in turn was expected to solidify a commitment to use and
maintain the facilities.
This consensus on the necessary components of a demand-driven planning model was largely silent
on the relative importance of these components. Were all three components of the model necessary
for success? Or were two out of three sufﬁcient? Was it possible to proceed in a sequential fashion,
implementing some parts of this planning model ﬁrst and following up with others at a later date?
There was an implicit assumption that, if other elements of the demand-driven planning model were
present and spare parts were available for purchase, communities would be prepared to assume
responsibility for the management and maintenance of their water systems without further post-
construction support (PCS).
The most controversial component of the policy advice in the demand-driven, community
management model has been to require that households pay a share of the capital costs and all of the
operation and maintenance costs of providing rural water services. In many cases rural households
without improved water sources are the very poorest and efforts to require such households to
pay something for services (or to target investments at communities that are collectively willing and able
to pay) have often disturbed both large multilateral donors and small non-governmental organizations
(NGOs) working in the sector. Small NGOs, such as church-based organizations and other charities,
especially want to help poor people in direct, tangible ways. Projects to provide improved water
supplies for poor rural households have often received strong support from both NGO staff and
their donor base and the argument that people should have to pay for such services has seemed
antithetical to the rationale and desire of many NGOs to be involved with such communities in the
Objections to charging poor rural households for improved water systems arise not only from a belief
that rich countries and individuals have a moral obligation to transfer ﬁnancial resources to poor
households in order to eliminate global inequities. Analysts have also argued that the provision of
improved water services has large positive health externalities and thus a traditional economic efﬁciency
D. Whittington et al. / Water Policy 11 (2009) 696–718698
criterion calls for the use of a Pigouvian subsidy to equate marginal social beneﬁts with marginal
A somewhat different line of argument in support of providing improved water services free of charge is
that poor households are caught in a vicious cycle of poor health, limited education and low economic
productivity and that improved water (and sanitation) services are one change that helps people to break
out of this “poverty trap” (Sachs, 2005). From this perspective, subsidized rural water projects are not only
equitable and morally obligatory, but they are also engines of economic growth. This argument for
simultaneously investing the health, education and infrastructure is exempliﬁed by the desire to meet the
multisectoral Millennium Development Goals by 2015
. Although there is no rigorous empirical evidence
that investments in water supply projects cause (or induce) economic growth in poor rural communities,
the “poverty trap” metaphor resonates strongly with NGOs and other donors uncomfortable with the cost
recovery component of the demand-driven, community management model
NGOs and multilateral donors have been less reluctant to shift non-pecuniary costs (such as time and
labor commitments) onto community members. The demand-driven, community management model
holds that much of the human resource costs of managing rural water projects should be transferred to
village water committees (VWCs). This proposal has been relatively uncontroversial.
Since the mid-1990s a number of donors, working with national and regional water resources ministries
in developing countries, have designed and implemented rural water supply programs that incorporated
one or more components of this demand-driven, community management model. Few of these programs
planned for systematic provision of PCS; community management was assumed to be feasible from a
technical perspective. Recently some have argued that it is unrealistic to expect that government can leave
rural communities to their own devices after a water project is completed and that for rural water supply
systems to be successful, communities need some PCS, such as follow-up training and technical
assistance visits by engineers (Kleemeier, 2000;Lockwood, 2002,2003).
There is little systematic evidence about how the demand-driven, community management model
(with or without PCS) is working in practice. Are the projects implemented using the demand-driven,
community management approach successful? Are the systems working? Are village water committees
able to have systems ﬁxed when they break? Do the committees function as planned? Are revenues being
collected? Are households satisﬁed with rural water supply projects? These are some of the questions we
address in this paper. Although the conventional wisdom in the water resources community still seems to
be that the rural water supply sector is prone to failure, much of the policy discussion remains focused on
how to increase donor support to expand rural water supply coverage (UNDP, 2006).
3. Research design
It is important to understand our initial research design in order to appreciate the strengths and
limitations of the ﬁndings presented in this paper. We wanted to investigate the effect of PCS on system
The empirical evidence for the existence of such positive externalities from improved water supplies is, however, surprisingly
limited. See Fewtrell et al. (2005) for a recent review of the effect of improved water, sanitation, and hygiene interventions on
This advice about the need for simultaneous investments across sectors has never been widely accepted by economists.
See, for example, Schumpeter (1939) and Hirschman (1958).
D. Whittington et al. / Water Policy 11 (2009) 696–718 699
sustainability when added to a well-designed demand-driven, community management rural water
. A gold standard research design would consist of a randomized controlled experiment
in which baseline conditions would be measured in villages, all of which had received an improved
water system as part of such a well-designed rural water supply program (Baker, 2000). Subsequently
PCS would be randomly assigned to treatment villages but not to control villages and then any
differences in treatment and control villages could be conﬁdently attributed to the PCS intervention.
We did not have the time or resources to assign PCS to treatment and control villages randomly and
then collect data on system performance before and after the treatment intervention
. The best we could
do was to try to ﬁnd situations where some villages had received PCS and other similar villages had
not. In order to establish the direction of causation between PCS and system performance, we needed to
ﬁnd well-designed demand driven programs in which some villages had received PCS “automatically”
(i.e. the treatment villages) and other villages had not (the control villages). In other words, we needed a
situation in which the treatment villages received PCS services without asking for them. Otherwise, one
could easily suspect that the treatment villages that demanded the PCS services (i.e. called in or sought
out PCS services) would be systematically different from the control villages.
In the design of the research project, we searched for rural water supply programs in developing
countries around the world that used a “state of the art” demand-driven, community management model,
but also used a “supply-driven” (automatic) PCS program to assist villages after construction. Demand-
driven, community managed rural water supply programs that have been in operation for several years
are not all that common; such programs with “supply-driven” PCS components are rare around the
world. We selected programs in Bolivia, Peru and Ghana that we believed came closest to meeting these
research design criteria.
In all three countries we included in our sample villages that had received improved water supply
projects 3– 12 years earlier as part of a demand-driven, community managed, donor-supported rural
water supply program. Approximately half of the villages in the sample were treatment villages and half
were controls. In all three countries we used secondary data to attempt to choose districts or regions
where the characteristics of villages and households in the treatment and control villages were likely to
In Bolivia both the treatment and control villages were part of PROSABAR (Proyecto de Saneamiento
´sico Rural), a World Bank-funded program that grew out of a water and sanitation program pilot
project implemented in the departamento of Potosı
´. We selected 99 PROSABAR villages in the
Chuquisaca and Cochabamba departamentos. Sample villages were located in the central highlands at
elevations of 1,800– 3,000 m; rainfall varied from 30– 69 cm per year.
Within each departamento, we tried to include approximately 25 communities that had received some
form of PCS and another 25 that had not. However, the determination of “treatment” and “control”
It did not seem interesting to study the effect of PCS in poorly designed programs because there is ample evidence in the
literature (e.g. Narayan-Parker, 1995;Sara & Katz, 1997) that these programs would have many failures and that PCS alone
could not salvage them.
Nor did we have the time or resources to measure baseline conditions in treatment and control villages and then wait for the
effects of the treatment (PCS) to unfold. Rather, we collected information from both treatment and control villages in the
current period, after some villages had received PCS. We were thus forced to ask people we interviewed today about conditions
in the past, both before and after their rural water system was constructed. This approach is not ideal, because people’s
memories are imperfect and some baseline data simply were not available to us.
D. Whittington et al. / Water Policy 11 (2009) 696–718700
communities was problematic because the municipal records regarding the provision of PCS were often
incorrect. Communities were later reclassiﬁed into the “treatment” or “control” group based on
interviews, reports and records related to PCS in each community. Across both departamentos, 34% of
the sample communities received some sort of PCS and 66% did not. The percentages varied for speciﬁc
types of support and by departamento, as discussed further below.
In Peru both treatment and control villages were located in the Cuzco region. The 43 treatment villages
were part of the Swiss-funded SANBASUR project; the 56 control villages were from the World Bank-
ﬁnanced FONCODES project. As in Bolivia, the sample villages in Peru were in the central highlands.
Elevations of villages were slightly higher than in Peru (2,500– 4,000 m); annual rainfall was about
70 cm. The average village size in Peru (588 people) was a third smaller than in Bolivia (870 people).
In both Bolivia and Peru the rural water systems, installed as part of the programs, were almost all
gravity-fed piped distribution systems with unmetered private connections. In some villages there were a
few public taps for unconnected households. These systems were, on average, seven years old at the time
of the ﬁeldwork. Capital costs in Bolivia were approximately US$80 per capita at the time of construction
(US$400 for an average household with ﬁve members); costs in Peru were probably comparable
In Ghana the treatment villages were selected from four districts in the Volta region and the control
villages from ﬁve districts in the Brong Ahafo region. The Ghana Community Water and Sanitation
Agency implemented the rural water supply programs in both regions. The treatment villages in Volta
were part of a Danida-funded program, which since 2003 included a PCS component called “monitoring
of operations and maintenance” (MOM); the control villages in Brong Ahafo were part of a World Bank-
funded rural water supply program. Both treatment and control villages were in lowland forests at
elevations of 74– 503 m. Annual rainfall was about 120 cm. The average age of the water systems in
Ghana was six years.
In Ghana, only villages with boreholes and non-mechanized public handpumps were included in the
sample. The results of recent contract awards in Brong Ahafo and Volta indicate that the total cost of
drilling a successful borehole and installing a handpump is typically in the range of US$10,000– 12,000.
The sample of both control and treatment villages was limited to communities that received no more
than two boreholes as part of the water supply program. This effectively also limited the size of the
villages; at the time of our ﬁeld visits in 2005 the population of sample villages ranged from 200 to 5,000
people. These selection criteria yielded a potential sample frame of 98 villages in the Volta region and
120 villages in the Brong Ahafo region. All 98 villages in the Volta region were selected and 104 of the
120 villages in the Brong Ahafo region were randomly selected.
Field conditions in Bolivia, Peru and Ghana presented us with some unanticipated challenges. Our
research design required that the water system in some villages be “successes” and in others “failures” in
order to have variation in our dependent variable. In fact, as described below, we found far fewer project
failures in either treatment or control villages than expected.
The complexity of PCS provision in the study settings posed another threat to our research design.
To establish the direction of causation between PCS and system performance, we tried to identify
PCS programs that were “supply-driven”, but in reality even supply-driven PCS programs often do not
work that way in practice. The Danida-funded MOM program in Ghana turned out to be the only true
supply-driven PCS program in our three study sites: villages in this program received quarterly visits from
We were unable to collect data on capital costs on the systems in Peru during the ﬁeldwork.
D. Whittington et al. / Water Policy 11 (2009) 696–718 701
environmental health assistants to monitor the technical, management and ﬁnancial status of the rural
water supply systems. It was more common to ﬁnd cases of communities seeking out PCS when their
water systems broke or when there was a management problem or conﬂict (i.e. “solicited” PCS)
. We also
found that help is often available to communities from more than just one “ofﬁcial” source. If a village
water committee cannot ﬁnd assistance in one place, there are often other places to turn to for help, such as
NGOs, nearby municipalities, or large commercial enterprises. Some villages may have the political clout
to obtain ﬁnancial assistance from a member of Parliament or a wealthy relative of a village resident
living abroad. Thus, even in PCS programs designed to be supply-driven, in practice much PCS is
demand-driven in the sense that many communities seek out help from wherever they can get it.
One implication of this complexity of PCS provision is that many control villages in our study sites had
various demand-driven forms of PCS available to them. Especially with the public handpump technology
used in Ghana, breakdowns are to be expected and thus repairs are necessary. Control villages (without
supply-driven PCS) must be able to obtain spare parts and mobilize the technical expertise necessary to
make repairs. Without these, all handpumps in villages in a rural water supply program (and some
gravity-fed distribution systems as well) will have broken down after only a few years. Our research
question was whether PCS services (both technical and non-technical) from higher levels of government
would lead to better performance of the village water system, although the baseline conditions in the
control villages was never no PCS at all. The control villages in any well-designed demand-driven,
community managed rural water supply program, at least initially, also have access to spare parts and to
some technical expertise (although for a variety of reasons control villages might not retain or mobilize
for use the training or procurement systems put in place at the time of project construction).
In summary, the implementation of our initial research design revealed the complexity of unraveling
the causation relationship between PCS and system performance. It also led us to study donor-funded
rural water supply projects and thus communities that were not randomly selected from either a global or
4. Data collection and proﬁle of sample villages
Fieldwork began in Peru in the summer of 2004, in Ghana in the autumn of 2004 and in Bolivia in early
2005. Data collection activities in each country were similar but not exactly the same. Generally a data
collection team spent one day conducting the ﬁeldwork in a village. During the course of the day, the team
held a group interview with members of the village water committee, interviewed the water system
operator or caretaker (and borehole attendant if applicable), conducted a focus group with women from a
diverse set of backgrounds, ages, ethnic and income groups and administered surveys to the heads of
household (or spouse) in approximately 25 households. In Bolivia we worked with leaders to draw
community maps and drew random samples of households. In Peru and Ghana, enumerators were
scattered more or less throughout a village and instructed to interview every ﬁfth household or some
other similar sampling rule. In Bolivia and Peru, interviews were conducted in either Spanish or
Given the choice of sending technical personnel to a community that has requested assistance and another community that is
due for a regularly scheduled “check up” visit, but is presumed to be doing ﬁne, it is natural for a manager of a “supply-driven”
PCS program to be inclined to direct resources to where the problem is.
D. Whittington et al. / Water Policy 11 (2009) 696–718702
Quechua; in Ghana in either Twi or Ewe. In addition to the survey and focus group information
collected by the ﬁeldwork teams, technical staff made an engineering assessment of the water supply
system in each village in the three countries. In Peru and Ghana, data collection also included a focus
group discussion with village leaders. In Ghana, the most heavily used borehole in the village was
observed for one day and data were collected on the quantity of water obtained. Most of the information
presented in this paper comes from interviews with village water committee members, water system
operators and households.
This information showed that the sample villages in Bolivia and Peru were small and remote; in
Ghana they were on average larger and more accessible. The majority of villages in Peru and Bolivia had
electricity and many households were connected to this service. In Ghana only 32% of the sample
villages were connected to the electricity grid. In all three countries the majority of households reported
that they were farmers. Average education levels were low; the typical respondent in all three countries
reported having “some” primary education. In Peru, 60% of the households interviewed reported annual
cash income of less than US$150 (vs. 42% in Bolivia). In Ghana, households reported median monthly
expenditures of about US$57. The average household had ﬁve members in Peru and Bolivia (vs. 6 in
Ghana). High percentages of households in Bolivia (86%) and Ghana (75%) reported trusting their
neighbors (vs. 51% in Peru).
5.1. Communities were involved in the pre-construction planning
Our interviews showed that community members felt that they had been involved in the pre-construction
planning of their water system. In all three countries over 90% of focus groups held with village leaders
and/or women reported that the community had been involved in tariff design. In approximately two-thirds
of the villages in Bolivia and Peru, people felt they had been involved in the choice of technology (vs. 42%
in Ghana). In slightly less than half of the villages in all three countries, people felt involved in decisions
about the site of projects (location of distribution lines in Peru and Bolivia, handpumps in Ghana). In all
three countries communities contributed 5– 10% of the capital costs of the project, but in many cases labor
and/or land contributions were allowed to substitute for cash. Overall our results conﬁrm that many of the
desired preconstruction elements of the demand-driven, community management model were
implemented in both treatment and control villages in all three countries.
5.2. Community water supply projects are still working
Based on our reading of the literature (Edig et al., 2002;Engel et al., 2003) and discussions with sector
professionals familiar with the local situation in all three countries, we expected to ﬁnd a substantial
minority—or perhaps even a majority—of the water systems in the villages in our sample to be
performing poorly or broken down, but this was not the case. In all three countries, in both treatment and
control villages the rural water supply projects were “working”. How one deﬁnes the performance of
rural water projects is somewhat more complicated than one might imagine, but in our case the deﬁnition
used did not affect our conclusion. As shown in Table 1, all the piped systems in Peru, and all but one of
the systems in Bolivia, were functioning at the time of our ﬁeld visit. Among the households included in
D. Whittington et al. / Water Policy 11 (2009) 696–718 703
our sample in Peru and Bolivia, 95% had operational taps at the time of our ﬁeld visit. In 55% of the
communities in Bolivia and 76% in Peru, 100% of the household taps were operational. In 90% of the
villages in Ghana, all project boreholes were still working.
This ﬁnding holds very good news for the rural water sector. The demand-driven, community
management model seems to be working, at least in the medium term. Not only were the rural water
Table 1. Proﬁle of village water systems and management practices.
Bolivia* Peru Ghana
Description of the system
Average years since project completion 7 7 6
Percentage of villages—private connections only 73 100 0
Percentage of villages—public taps only 4 0 100
Percentage of villages—private connections and public taps 23 0 0
Percentage of villages with only one project handpump N/A N/A 50
Status of the system
Percentage of households with functioning taps 95 95 N/A
Percentage of villages with all taps functioning 54 74 N/A
Percentage of villages where all project handpumps are working N/A N/A 89
Percentage of villages with functioning systems, which had reported
a breakdown over last six months
55 55 57
Average days to repair the system
(for villages that had experienced a breakdown)
1–2 5 18
Percentage of villages where committees regularly hold meetings
with the community
86 81 72
Percentage of villages where committee members are elected 95 63 42
Percentage of villages where committee members are appointed 3 15 43
Median number of women on the committee 0 0 3
Percentage of villages with no caretaker/operator 3 2 18
Percentage of villages with paid caretaker/operator
(in villages with a caretaker)
70 57 1
Cost recovery mechanisms
Pay-by-the bucket or volumetric tariff 2 0 39
Fixed monthly fee 89 82 54
Fees vary by household size 0 0 7
Irregular collections 0 7 16
No revenue collection 9 11 13
Percentage of households in full sample who use the system that
reported paying for water
87 77 71
Median monthly expenditure for water reported
among households that pay for water (US$)
0.55 0.30 N/A
Percentage of committees reporting that household
collections cover operating costs
N/A 50 51
Percentage of committees reporting that household
collections cover minor repairs
N/A 80 65
Percentage of committees reporting that household
collections cover major repairs
N/A 12 30
*88% of systems in Bolivia were gravity only; the others used pumps.
D. Whittington et al. / Water Policy 11 (2009) 696–718704
systems producing water (i.e. had not broken down), but almost all the households in these communities
were obtaining at least some of their water from the systems. In Bolivia, 100% of the households
interviewed reported using water from the improved water system; in Peru it was 95% and in Ghana
97%. In Ghana, our estimates of the amount of water collected by households from boreholes ranged
from 34 liters per day in Volta to 24 liters per day in Brong Ahafo. These levels of per capita water use
are quite high for rural areas of Africa when people carry water from a source outside their home (White
et al., 1972;Mu et al., 1990;Katui-Kafui, 2002) and indicate that these borehole projects have
succeeded in terms of supplying relatively large quantities of water for household use.
In all three countries, the vast majority of VWCs are functioning as planned. They hold regular
meetings. Many of the VWC members were elected and, in Ghana, many of these were women
. In Bolivia and Peru every community had a system operator who was responsible for the
operation and maintenance of the water system. In Ghana, 82% of the communities still had a caretaker
for the handpump(s). The systems in all three countries do occasionally break down, but in the majority
of cases the VWCs are able to arrange for repairs. The majority of villages in all three counties reported
one or more breakdowns in the last six months, but in Bolivia this was typically ﬁxed in 1 – 2 days. In
Peru, breakdowns were ﬁxed on average in ﬁve days and in Ghana in 18 days
In Ghana most villages in the sample had functioning VWCs: only 3% of VWCs in the study villages
in Volta and 7% in Brong Ahafo had been disbanded or relieved of their duties. Another 5% of VWCs
were inactive or dormant. There are a variety of different explanations for these instances of VWC
failure or inactivity. In some cases the committee had stopped work owing to conﬂicts with the
community or village leaders (usually over revenue collection, the use of collected revenues, or
unsuccessful repairs). In others, the committee was dormant because there was “no work to do” (the
borehole had either not broken down or had not functioned in a long time). In a few villages another
village-level institution had assumed the responsibility for the water system.
In all three countries the water systems were working, communities were able to make repairs and as a
result, levels of household satisfaction were very high in most villages. On average, in Bolivia, 83% of
households in each village reported being “satisﬁed” or “very satisﬁed” with their system’s operation and
maintenance (O&M) regime and 78% with the performance of their VWC. In Peru, 61% of households
reported that they were satisﬁed overall with the improved water system. In Ghana, 88% of households
interviewed reported that they were satisﬁed with the repair and maintenance services of their water
system and over 80% of the women’s focus groups said they were satisﬁed with the systems
5.3. Households still using unprotected sources
But there are also some troubling ﬁndings. Although almost all households reported using the new
water system, for some households this was not their only water source. Especially in Ghana, 38% of
households still reported using water from unprotected sources (e.g. springs, river, open wells) for
drinking and/or cooking. The number of households using unprotected sources for these purposes was
This does not necessarily mean that the women are active committee members (see evidence from India in Prokopy, 2004).
The main reason that repairs take longer in Ghana is that parts for boreholes must be obtained from outside the villages.
In Bolivia and Peru many of the repairs can be made with parts that communities have on hand.
Dissatisfaction in Ghana was primarily concentrated in villages where the handpumps were no longer working or had always
had problems (e.g. salty water or low pressure in the dry season).
D. Whittington et al. / Water Policy 11 (2009) 696–718 705
lower in Peru (21%) and Bolivia (23%), but still worrying. We do not have information on the health
consequences for the substantial minorities of the population in our sample villages which continue to
use traditional water sources, but we speculate that until households obtain their drinking and cooking
water exclusively from improved sources, the health beneﬁts of the investments in improved sources will
not be fully realized and any prospects for breaking out of a rural “poverty trap” will be reduced.
5.4. Households pay little for improved services
Another worrying ﬁnding is that rural households in the sample villages are paying very little for the
improved water services and, as a result, the ﬁnances of many VWCs are in poor shape. As noted, these
rural water supply programs were not designed for communities to recover the capital costs of
construction or to provide for capital replacement or expansion. The cost recovery objective was simply
to collect sufﬁcient revenues from users on an ongoing basis to pay operation and maintenance costs. But
a substantial minority of villages in our study is not achieving even this modest objective.
In both Bolivia and Peru almost all villages charged households a very modest ﬁxed monthly fee
for service. In Bolivia 87% of households and 77% of households in Peru reported paying for water, but the
median monthly expenditures were only US$0.25 and $0.66, respectively. In Bolivia the monthly charges
were not only low, but 27% of communities had actually lowered their tariffs since operation began. In
Peru, we estimate that slightly less than half of the communities manage to recover their operating costs.
In Ghana, 13% of VWCs say that they do not collect any money from households. When we asked
households (in contrast to VWC members), in 23% of the villages we found no households interviewed
that reported paying for water. Only 71% of the villages in Ghana have any regular payment system for
households (either pay-by-the bucket or ﬁxed monthly charge, as opposed to an irregular system like
collecting money from households when funds are needed for repairs). In villages that used ﬁxed
monthly fees, the most common rates were US$0.11 and US$0.22 per month. In villages using pay-by-
the-bucket, the most common charges per 20-liter container were US$0.01 or less.
Among the VWCs in Ghana that did collect revenue from households, those in Volta reported
collecting an average of US$169 annually from households (versus US$173 in Brong Ahafo). Revenues
of this magnitude should be sufﬁcient to pay for routine operations and maintenance, but not major
. However, the range of revenue collections reported by these VWCs was very wide, with some
committees saying they collected less than US$1 from all households during the entire year and others
reporting household contributions above US$2000. Nearly three-quarters of the VWCs in Ghana that
reported charging households for water (regular or irregular payment system) felt that they collected
enough money to pay for the cost of operations. Eighty nine percent said that they could pay for minor
repairs with the money collected from households, but only 41% of the VWCs that were charging
households for water said that they could pay for major repairs.
A 1994 study of the Afridev handpump in Ghana’s northern region found that the average annual cost to a community of
ﬁxing common problems, such as rod breakages, plus the cost of replacing fast-wearing parts like bobbins, U-seals, O-rings and
bearings, would be about US$60 (Osafo-Yeboah, 1994). UNDEP’s International Environmental Technology Centre (2009) puts
the expected operation and maintenance cost for communities with handpumps serving 200 to 300 people at between $0.26 to
$0.52 per capita per year. VWCs in our sample reported spending about US$100 annually on repairs. None of these estimates
include the real resource costs associated with the time invested by the VDC, the caretakers, or borehole attendants.
D. Whittington et al. / Water Policy 11 (2009) 696–718706
5.5. Villages use post-construction support
Despite the problems many communities have charging in households for water services, the majority
of communities in all three countries are managing to keep their improved water systems functioning.
Even communities that are not collecting sufﬁcient revenues to pay for operation and maintenance costs
are ﬁnding the resources they need to ﬁx their systems when they break. Especially in Ghana, many of
these resources come from outside the community (Table 2).
Nearly half of the communities in all three countries have received additional training for their water
system operators or caretakers since construction. Some villages have received help with non-technical
matters, such as billing or disputes over water sources. When water systems break, system operators seek
out spare parts and, if necessary, outside technical expertise to make repairs. Few VWCs keep sufﬁcient
cash on hand to pay for major repairs. Nonetheless, they seem able to ﬁnd the funds for repairs
somewhere, be it through one-time special assessments of villagers, through grants from outsiders, or in
the form of free parts or repair services. In some cases, the caretakers or VWCs turn to “middle men” to
help identify and obtain the resources they need. In Ghana, for example, the District Water and
Sanitation Teams (DWST) and the environmental health assistants involved in the MOM program have
helped communities ﬁnd technical assistance and spare parts. But other actors help as well. One of the
striking ﬁndings from our ﬁeld activities was the pervasive presence of NGOs and church organizations
in PCS activities.
Many NGOs are providing both supply-driven and demand-driven PCS. In Bolivia, NGOs like Plan
International and CARE have taken on increasingly programmatic roles in the rural water sector in the
sample villages as the role of government has diminished. In recent years they have largely assumed
responsibility for PCS. In Peru the “prime contractor” NGO (Sanbasur) assisted communities in ﬁlling
the gap between the revenues raised and funds needed for repairs by putting such communities in touch
with other partner NGOs or with municipal governments who could provide ﬁnancial and other
assistance. In Ghana, fully 16% of the sample villages have received grants for repairs and/or major
rehabilitation from outside sources such as the Church of the Latter Day Saints (Table 2). The Mormons
and perhaps other NGOs appear to have worked with the DWSTs to identify villages that are
experiencing problems and then to help ﬁnance the repairs. The DWSTs not only lack funds to monitor
conditions proactively in villages, but also are instructed by policy guidelines not to make or fund repairs
Table 2. Proﬁle of PCS activities.
Percentage of villages that received .... after completion of project construction Ghana (%) Peru (%) Bolivia (%)
Visits from external organization(s) to assist with maintenance or repairs 52 14 22
Visits from external organization(s) to assist with accounting, tariffs, etc. 33 6 13
Technical training for the system operator 34 49 41
Free repairs 21 N/A N/A
Written manuals or other materials 37 25 30
Help with ﬁnding or receiving spare parts 45 7 11
Grants from outside sources for repairs, new construction, system rehabilitation,
capacity expansion, or other assistance
16 3 8
Percentage of households visited by external agencies to discuss use of water
30 25 N/A
D. Whittington et al. / Water Policy 11 (2009) 696–718 707
5.6. Factors are associated with sustainability and satisfaction
Communities are making use of a wide-variety of government- and NGO-provided PCS services to
keep their systems working, some of which are provided at their request (“solicited PCS”) and others
(“supply-driven”) at the initiative of government or NGO or church organizations. Our cross-sectional
research design and the character of PCS in the three countries make it very difﬁcult to draw deﬁnitive
conclusions about the contribution of different forms of PCS to system sustainability. Nonetheless, we
used multivariate models to investigate the factors (including PCS) that were associated with whether or
not the village’s improved water systems were working and whether households were satisﬁed with the
service they received.
For our analysis of technical sustainability, we explored a variety of deﬁnitions of our dependent
variable. The one we choose to report for Bolivia and Peru is simply whether or not 100% of the
household connections were functioning in the village at the time of our visit. For Ghana, our dependent
variable is whether or not all the project handpumps and boreholes were operational (supplying water) at
the time of our visit. Table 3 presents the means, medians and standard deviations of our independent
variables. Table 4 reports the results of the model for each of the three countries.
Some of the factors positively associated with good system performance are as expected. In Bolivia
and Ghana (but not in Peru) electricity coverage was positively associated with good system
performance, which we think is likely to be a wealth effect. In Peru the age of the water system was
negatively correlated with system performance (statistically signiﬁcant at the 1% level), but this may be
because the sample of communities in Peru included some water systems that were completed more
recently than in Peru or Ghana. There is no association between age and system performance in Bolivia
or Ghana—which we interpret as further evidence that the demand-driven, community management
model is working as hoped in the medium-term.
In none of the three countries was there a statistically signiﬁcant association between a village
receiving a technical PCS visit (to help with repairs or maintenance) and having a working water system.
Post-construction technical training of system operators or caretakers was, however, positively
associated with system performance in both Ghana and Bolivia.
In Bolivia, water systems in the Chuquisaca region were more likely to be working than in the
Cochabamba region. In Peru, projects in the SANBASUR program were more likely to be working than
those in the FONCODES program. In Ghana, there was no statistically signiﬁcant association between
region (Brong Ahafo vs. Volta) and system performance. This ﬁnding suggests that a labor-intensive
supply-driven PCS program like MOM (quarterly audits of the technical and ﬁnancial function of the
water supply systems by environmental health assistants) does not increase the technical sustainability
of handpump systems in a setting like rural Ghana where communities have access (if requested) to
many others forms of PCS
Because the technology in Ghana (handpumps) was different from that in Bolivia and Peru (gravity-
fed systems with household connections), the Ghana model has three independent variables not included
in the Bolivia and Peru models: (1) whether the village had only one borehole, (2) population per
borehole and (3) whether the village had an unprotected source that always has water during the dry
We did not investigate whether the MOM program improved hygiene, water use habits, or cleanliness of the handpump sites,
all of which would be other expected beneﬁts of regular visits by environmental health assistants.
D. Whittington et al. / Water Policy 11 (2009) 696–718708
Table 3. Summary statistics (mean and standard deviation) of variables used in the multivariate models.
Variable name Variable deﬁnition Bolivia (n¼77) Peru (n¼99) Ghana (n¼175)
Satisfaction % of households who report being satisﬁed with:
Maintenance and operations of piped water system
Mean: 83 Mean: 70 Mean: 0.87
Maintenance and repair (Peru)
Std. dev: 20 Std. dev: 19 Std dev: 0.15
Preventative maintenance and repair service (Ghana)
Median: 90 Median: 72 Median: 0.92
System working 1 ¼All sampled taps in the village are functioning
(Bolivia and Peru)
Mean: 0.59 Mean: 0.75 Mean: 0.90
1¼All project handpumps in the village are
Std. dev.: 0.50 Std. dev: 0.44 Std. Dev: 0.30
Median: 1 Median:1 Median:1
System age Number of years since system began operation
(Bolivia and Peru)
Mean: 7.0 Mean: 6.7 Mean: 6.0
Number of years since handpumps were installed
Std. dev.: 1.2
Std. dev: 2.3 Std. dev: 0.8Median: 7.0
Median: 7.0 Median: 6
Number of handpumps 1 ¼village received only one hand pump (Ghana) N/A N/A Mean: 0.51
Std. dev: 0.50
Population per handpump Population per handpump installed by project
(100s of persons)
N/A N/A Mean: 6.43
Std. dev: 6.38
Electricity coverage Percentage of households interviewed with electricity Mean: 39.0 Mean: 60.5 Mean: 13.69
Std. dev.: 41.2 Std. dev: 39.5 Std. dev: 23.49
Median: 14.5 Median: 80 Median: 0
Remoteness Distance in kilometers to:
Mean: 44 Mean: 58 Mean: 19
...paved road (Peru)
Std. dev.: 57 Std. dev: 85 Std. dev: 18
...area mechanic (Ghana)
Median: 24 Median: 15 Median: 15
Trust of neighbors Percentage of households interviewed who say they
trust their neighbors
Mean: 83 Mean: 56 Mean: 74
Std. dev.: 18 Std. dev: 18 Std. dev: 14
Median: 88 Median: 56 Median: 76
Reliable unprotected alternative source Village has unprotected source that always has
water during the dry season within 1 km of
N/A N/A Mean: 0.21
Std. dev: 0.41
D. Whittington et al. / Water Policy 11 (2009) 696–718 709
Table 3. Continued
Variable name Variable deﬁnition Bolivia (n¼77) Peru (n¼99) Ghana (n¼175)
Technical training 1 ¼During the post-construction period, water
system operator or village caretaker has received
Mean: 0.28 Mean: 0.36 Mean: 0.39
Std. dev.: 0.45 Std. dev: 0.48 Std. dev: 0.49
Median: 0 Median: 0 Median: 0
Technical PCS visit 1 ¼During the post-construction period:
...received $1 unsolicited technically-oriented visit
Mean: 0.20 Mean: 0.07 Mean: 0.19
...received $1 unsolicited visit to assist with
Median: 0 Median: 0 Median: 0
...received $1 unsolicited free repair (Ghana)
Std. dev.: 0.40 Std. dev: 0.26 Std dev: 0.36
Financial or managerial PCS
1¼During the post-construction period: Mean: 0.10 Mean: 0.04 Mean: 0.29
...received $1 unsolicited non-technically oriented
Std. dev.: 0.30 Std. dev: 0.2 Std. dev: 0.46
...received $1visit to assist with ﬁnancial or
Median: 0 Median: 0 Median: 0
Regional or project identiﬁers Bolivia: Mean: 0.50 Mean: 0.45 Mean: 0.48
1¼Cochabamba region Std. dev.: 0.50 Std. dev: 0.5 Std. dev: 0.50
0¼Chuquisaca region Median: 0.50 Median: 0 Median: 0
0¼Brong Ahafo region
Theoretically all villages in Volta should have received assistance with ﬁnancial and managerial matters through the MOM program, but not all VWCs
perceived the MOM audits as such.
D. Whittington et al. / Water Policy 11 (2009) 696–718710
Table 4. Factors associated with “system working”.
Bolivia Peru Ghana
Dependent variable Household taps
(1 ¼100% working,
(1 ¼100% working,
in the village are
(SE: 0.01) (SE: 0.003) (SE: 0.016)
Odds ratio: 1.01 Odds ratio: 1.00 Odds ratio: 1.01
Electricity coverage 0.03* 0.013 0.08
(0.01) (0.009) (0.035)
1.03 1.01 1.09
Technical PCS visit 0.17 1.172 21.16
(0.83) (1.226) (0.82)
1.19 3.23 0.31
Financial or managerial
0.43 21.047 20.04
(1.11) (1.661) (0.86)
1.54 0.35 0.96
Technical training 1.16‡ 20.390 1.48†
(0.69) (0.595) (0.71)
3.19 0.68 4.43
System age 20.42 20.346* 0.46
(0.27) (0.147) (0.49)
0.66 0.71 1.58
Trust of neighbors 20.04 1.368 0.04
(0.02) (2.038) (0.03)
0.96 3.93 1.03
Region or program
(0.97) (0.78) (0.79)
0.05 3.99 2.16
One borehole 1.99*
(3.23) (2.11) (4.11)
value 0.23 0.14 0.30
77 86 175
Signiﬁcant at 0.01 level.
Signiﬁcant at 0.05 level.
Signiﬁcant at 0.10 level.
D. Whittington et al. / Water Policy 11 (2009) 696–718 711
season within 1 km of the village. All three were statistically signiﬁcant factors associated with system
If a village had only one borehole, it was more likely to be working. We interpret this to mean that the
VWC makes more effort (and is under more community pressure) to keep the borehole working if there
is only one in the village. We interpret population per borehole as a measure of the pressure on the
resource and a measure of crowding. It is negatively associated with system performance, which we
speculate means that more intensive use of boreholes in these communities leads to the need for more
difﬁcult and expensive repairs, or that households value the handpump less when it must be shared with
more households and thus put less pressure on the VWC to keep it working. Our interpretation of the
negative association between having a reliable alternative source and a functioning water system is
similar—households have less need for the improved water source when there is a reliable alternative
source nearby, value the new source less and put less pressure on the VWC to keep the handpump
working (World Bank Water Demand Research Team, 1993).
IWe used similar multivariate models to investigate the factors that were associated with whether
or not households in a community said that they were satisﬁed with different aspects of their
improved water system. Again, we explored a variety of deﬁnitions of our dependent variable. For
Bolivia and Peru we chose the percentage of households in the village that reported they were
satisﬁed with operation and maintenance of the water system. In Ghana we used a similar deﬁnition:
the percentage of households in the village that reported they were satisﬁed with repairs and
maintenance of the water system
.Table 5 reports the results of this “satisfaction” model for each of
the three countries.
The striking result in the Bolivia satisfaction model is that the percentage of households in a village
that is satisﬁed is an average of 15 percentage points higher if the village has received a PCS visit that
provided ﬁnancial or managerial (but not technical) assistance. This effect is large and statistically
signiﬁcant; it is robust to model speciﬁcation and the deﬁnition of the dependent variable (Davis et al.,
2008). In Peru, the percentage of households living in the village that is satisﬁed is lower if the water
system in the village is older, but this is not the case in Bolivia and Ghana.
In the Ghana model, there are a number of independent variables associated with the percentage
of households in a village that report they are satisﬁed. First, if a village has received a technical
PCS visit, the percentage of household who say they are satisﬁed is lower. We interpret this as
evidence that the technical PCS visit was not exogenous and that villages receiving technical PCS
may already have been in trouble. However, technical training was positively associated with
satisfaction. Consistent with the ﬁnding from Bolivia, if a village had received a PCS visit providing
managerial or ﬁnancial assistance, a higher percentage of households reported being satisﬁed.
Villages in which a higher percentage of households reported that they trusted their neighbors also
had a higher percentage of households who were satisﬁed with the repairs and maintenance of their
water system. Finally, in Ghana villages with large populations per borehole, a lower percentage of
households were satisﬁed, which we interpret as consistent with the results for population per
borehole in Table 4.
We chose to look at satisfaction with repair, maintenance and operations because these are within the control of the
community. Satisfaction with water quality or overall satisfaction with the system may depend on construction and water-
resource related factors over which communities have little control once construction has been completed.
D. Whittington et al. / Water Policy 11 (2009) 696–718712
Table 5. Factors associated with households’ satisfaction with O&M repair services.
Bolivia Peru Ghana
Dependent variable HH satisfaction with
O&M (% satisﬁed)
HH satisfaction with
O&M (% satisﬁed)
HH satisfaction with repair
and maintenance (% satisﬁed)
(SE ¼0.04) (SE ¼0.00) (SE ¼0.00)
Electricity coverage 0.18* 0.00 0.00
(0.07) (0.01) (0.00)
Technical PCS visit 23.21 0.07 20.97
(5.59) (0.08) (0.031)
Financial or managerial PCS
14.64* 20.02 0.06*
(7.32) (0.12) (0.03)
Technical training 1.94 0.06 0.06
(5.06) (0.04) (0.02)
System age 22.17 20.02* 20.001
(1.95) (0.01) (0.015)
Trust of neighbors 0.74
(0.14) (0.14) (0.001)
Region or program dummy
(5.78) (0.05) (0.028)
One borehole 20.012
Population per borehole 20.004*
Reliable alternative source 20.03
(16.35) (0.12) (0.13)
value 0.30 0.08 0.17
Number of observations 77 89 175
Signiﬁcant at 0.05 level.
Signiﬁcant at 0.01 level.
Signiﬁcant at 0.10 level.
D. Whittington et al. / Water Policy 11 (2009) 696–718 713
We are unaware of other reports from the ﬁeld of such encouraging ﬁndings from large numbers of
rural communities in different countries. If these ﬁndings turn out to be as robust as we hope, it seems
that the demand-driven, community management model, coupled with access to spare parts and some
technical expertise, has come a long way towards unraveling the puzzle of how to best design and
implement rural water supply programs in developing countries.
Our conclusions on the relationship between PCS and sustainability are more tentative and merit
further investigation in other ﬁeld sites. The communities in our study solicit and use a wide range of
PCS services that are available to them. Nonetheless, we ﬁnd no evidence that the provision of free
repairs or free technical assistance, or that implementing an intensive supply-driven PCS program like
MOM in Ghana, are positively associated with improved technical sustainability or increased household
satisfaction. This supports the wisdom of the original conception of the demand-driven community
management model—that communities can and should take full responsibility for their systems. The
non-solicited PCS activities that appear most promising from this study are those that help communities
renew and further develop their capacities: post-construction training for system operations and non-
technical support visits to help VWCs with administrative functions or water use disputes.
Our ﬁndings also present some challenges for the sustainability of investments in the rural water
sector. One overarching issue is that even the communities we studied where cost recovery systems seem
to be meeting program objectives (i.e. villages pay 5– 10% of capital costs and collect tariffs to cover
operation, maintenance and repairs) are not moving toward a ﬁnancially sustainable future in which they
can (1) replace infrastructure when it reaches the end of its economic life, or (2) expand system capacity
to accommodate population and economic growth. Donor-funded rural water supply programs have
been structured as one-time investment programs, designed to meet only the immediate needs of rural
communities. This means that the moral obligation assumed by higher level government and donors is
not over. The current ﬁnancing system ensures that these communities will keep returning for capital
subsidies, just as some are doing now for repairs.
Some might argue that this is not a problem, that as long as poor people need help they should get it.
But the indirect consequences of this capital ﬁnancing model need to be carefully considered. In Ghana,
part of the reason some households continue to rely on traditional sources appears to be that capital
subsidies were spread too thinly and that an insufﬁcient number of boreholes were installed to serve a
growing population. In Bolivia, one consequence of a per-capita cap on capital expenditure (designed to
provide a disincentive for communities to ask for capital-intensive, perhaps inappropriate, facilities)
seems to have been that some communities restrict their service boundaries and leave households on the
periphery without piped services. Although these unconnected households were provided with wells,
they may prefer in the future to upgrade to the level of service enjoyed by their neighbors. Expanding
coverage in these Bolivian systems will be complicated by the fact that the water sources in many
villages do not have sufﬁcient water to serve additional people. Two-thirds of the women who
participated in focus group discussions felt that the quantity of water provided by PROSABAR systems
was insufﬁcient or “just enough to meet community needs”. Approximately 20% of the PROSABAR
communities studied have experienced decreases in the quantity of water supplied during the dry season
since their system was constructed.
In large municipalities, new water systems are routinely designed with excess capacity in both the
distribution system and the water source to provide for growing populations. But everywhere in the
D. Whittington et al. / Water Policy 11 (2009) 696–718714
rural water sector, capital subsidies are limited and excess system capacity is one of the ﬁrst
casualties. Moreover, few demand-driven rural water supply programs have incorporated a systematic
approach for providing follow-up capital subsidies to villages that have outgrown their current
systems or want to upgrade to a higher-level of service. Some of the communities in Ghana need to
plan for piped distribution systems that can support new businesses and other enterprises and the
current model for the provision of subsidized boreholes will not make this transition easy. Without the
option of gravity-fed distribution systems such as in Peru and Bolivia, the Ghanaian communities will
have higher O&M expenses. They will also need to plan for expenditures for system expansion. For a
village to do this on its own will require a cost recovery system that can generate a much higher and
more regular stream of revenues.
This brings us to a second major challenge for water sector professionals that is brought to light by this
study: why is it that the VWCs in a signiﬁcant number of villages are not collecting tariffs at all, or
collect insufﬁcient revenue from households to cover the ﬁnancial costs of major repairs, much less the
costs of system expansion or capital replacement? One possible explanation is that the initial capital
contribution that villagers made was not an adequate “demand ﬁlter”: making a nominal contribution to
capital costs (5– 10% of capital cost through cash or in-kind contributions) was not enough to ensure that
households in the recipient communities would be willing to pay the full ﬁnancial (and non-pecuniary)
costs of operating and repairing their new systems. We cannot rule out a possible link between low
capital contributions and poorly performing tariff collection systems, but neither do we have evidence
that increasing the initial capital contribution would lead to better cost recovery from households.
Rather, our ﬁndings suggest three principal reasons that VWCs are unable or unwilling to charge
First, generating substantial cash balances creates tough problems for the VWCs. These rural
communities do not have access to a convenient, secure banking system for the management of cash.
The median distance to the nearest municipality among the Bolivia villages in our sample was 24 km and
the average Peruvian village was 15 km from a paved road. Villages in Ghana are on average located
15 km from the urban centers where the area mechanics live and work. Moreover, many households have
little cash to spare and cash ﬂow is irregular and highly seasonal. Households are also often distrustful of
the accounting and security of cash balances and VWC members may be distrustful of each other or not
want the responsibility of securing cash.
Second, when VWCs do accumulate cash balances, villages often want to spend these monies on other
development projects. There is thus little incentive for VWCs to attempt to generate the funds necessary
for major repairs to the water system if they will “lose” them anyway. In such a situation, it makes sense
just to wait and try to raise the funds when the need arises. For all these reasons, life is much simpler for
members of VWCs if cash is only sufﬁcient to pay for minor O&M costs or is only collected at the
moment funds are needed.
Third, VWCs may well be correct to believe that future capital and repair subsidies will be
forthcoming from donors, NGOs and higher levels of government when they are needed. Not only was
the vast majority of the capital for these projects provided at no cost to the communities at the time of
construction, but a signiﬁcant number of VWCs in our sample have successfully found ways to insulate
households from the cost of repairs to the water systems. They have obtained donations, free spare parts
and free repairs from a wide variety of NGOs, church organizations, private individuals and companies
and even local governments. Herein lies a third major challenge for rural water supply policy: does the
sector’s current capital ﬁnancing model—and the post-construction activities of these NGOs and other
D. Whittington et al. / Water Policy 11 (2009) 696–718 715
actors—create a moral hazard that will undermine the principle of community self-reliance in the post-
In Ghana, the fact that 1 in 6 of the sample villages had received grants from outside sources after the
construction of the project may not seem like much, but this means that almost all VWCs would know
that NGOs and others are active and nearby. It may seem like a reasonable bet to wait until major repairs
are needed and see if an NGO might provide the cash infusion required. Moreover, an effort by a VWC
to establish some kind of sinking fund to make major repairs and replace capital at some future date may
make the community “less needy” to the NGO and actually preclude the community from receiving such
support. Indeed, small towns in the United States face similar disincentives to ﬁnancing their own
capacity expansion and system rehabilitation.
From the perspective of the NGO, repairing a handpump or ﬁxing a broken transmission line for a
piped distribution system may well seem like an ideal project. With a relatively small amount of
incremental funds, the NGO can reasonably claim to its donor base that all beneﬁts of the infrastructure
are due to its involvement, because without the incremental investment the system would have remained
broken. NGOs (and other donors) are especially attracted to such opportunities where their funds have
great “leverage”. But this funding strategy raises two important questions. First, would the community
have managed to raise funds locally and made the repair if the NGO had not been standing by ready to
step in? Second, if all the credit for the infrastructure goes to the “last investor” in, who is going to be
willing to continue making the capital infusions necessary to replace the aging capital stock, that is, to do
the “heavy lifting” that is required under this capital ﬁnancing model? Will higher levels of government
and donors step into these rural communities 5 or 10 years down the road when these systems are fully
depreciated and replace the capital that NGOs have kept running?
The present situation in the rural communities in our sample is not ﬁnancially sustainable without new
infusions of capital investments in the relatively near future—both to replace existing infrastructure and
to provide for economic growth. The moral hazard from the active involvement of NGOs, religious
groups and other non-state actors in the rural water sector is likely to prove to be an important factor
undermining cost recovery efforts and may discourage communities from making their own investments
in water infrastructure to support economic growth.
Long-term ﬁnancial sustainability requires a different policy model. Communities do want and need
help, but this assistance should not perpetuate their dependency on NGOs or higher levels of government
for limited capital subsidies that lock communities into infrastructure systems that are not suited for
achieving economic development or for accommodating growing populations. Nor should it undermine
local initiatives to pay for higher levels of infrastructure or infrastructure expansion. The coordination of
the policies of NGOs with government and with each other seems especially important and worthy of
future research. The involvement of NGOs in the sector has proven important for fostering policy
innovation, serving the very poorest and helping communities ﬁnd the resources they need to keep their
water systems running. But as suggested by our ﬁndings, NGOs can also create moral hazard problems
that may ultimately undermine rural economic development. One important role for NGOs in the future
could be as a catalyst for PCS, rather than as dispenser of capital subsidies for communities that cannot
manage to repair their water projects.
In summary, the demand-driven, community management planning model has come a long way
towards ﬁnding the key to success in the rural water sector. The next frontier seems to be the design of a
policy framework that will enable communities to handle the twin challenges of system rehabilitation
D. Whittington et al. / Water Policy 11 (2009) 696–718716
This research was sponsored by the Netherlands-World Bank Partnership. The research would not
have been possible without our collaborators in Peru, Bolivia and Ghana. We would especially like to
thank our co-authors and consultants on the individual country studies, including Alfonso Alvestegui
(Consultant, Bolivia), Betty Soto (IMPACTO, Bolivia), Gloria Liza
´rraga (IMPACTO, Bolivia), Bernard
Akanbang (TREND, Ghana), Benedict Tuffuor (TREND, Ghana), Eugene Larbi (TREND, Ghana),
Brenda Bucheli (PACT Peru), Jorge Izaguirre (PACT Peru).
We also appreciate the helpful comments of Rob Chase, Vijayendra Rao, Jennifer Sara, Robert Roche,
Donald T. Lauria and Marc Jeuland. All opinions expressed in the paper are those of the authors and
should not be attributed to these reviewers, the Netherlands Government, the Water & Sanitation
Program, or the World Bank.
Baker, J. (2000). Evaluating the Impact of Development Projects on Poverty: A Handbook for Practitioners. The World
Bank, Washington, DC.
Briscoe, J. & DeFerranti, D. (1988). Water for Rural Communities: Helping People Help Themselves. The World Bank,
Churchill, A., De Ferranti, D., Roche, R., Tager, C., Walters, A. & Yazer, A. (1987). Rural Water Supply and Sanitation:
Time for a Change. World Bank Discussion Paper 18. Washington, DC, World Bank.
Davis, J., Lukacs, H., Jeuland, M., Alvestegui, A., Sotto, B., Lizarraga, G., Bakalian, A. & Wakeman, W. (2008). Sustaining the
beneﬁts of rural water supply investments: experience from Cochabamba and Chuquisaca, Bolivia.Water Resources
van Edig, A., Engel, S. & Laube, W. (2002). Ghana’s water institutions in the process of reforms: from the international to the
local level. In Reforming Institutions for Sustainable Water Management. Neubert, S., Scheumann, W. & van Edig, A. (eds).
German Development Institute (DIE), Bonn.
Engel, S., Iskandarani, M. & Useche, M. (2003). Improved water supply in the Ghanaian Volta Basin: Who uses it and who
participates in community decision making. Paper presented at the Conference of the International Association of
Agricultural Economists (IAAE), Durban, South Africa, August 2003. Forthcoming as IFPRI-EPTD Discussion Paper,
International Food Policy Research Institute, Washington, DC.
Fewtrell, L., Kaufmann, R. B., Kay, D., Enanoria, W., Haller, L. & Colford, J. M. Jr (2005). Water, sanitation and hygiene
interventions to reduce diarrhoea in less developed countries: a systematic review and meta analysis.Lancet: Infectious
Diseases,5(1), 42– 52.
Hirschman, A. (1958). The Strategy of Economic Development. Yale University Press, New Haven.
Katui-Kafui, M. (2002). Drawers of Water II: Kenya Country Study, International Institute for Environment and
Kleemeier, E. (2000). The impact of participation on sustainability: an analysis of the Malawi rural piped scheme program.
World Development,28(5), 929– 944.
Komives, K., Akanbang, B., Wakeman, W., Thorsten, R., Tuffuor, B., Bakalian, A., Larbi, E. & Whittington, D. (2007).
Community management of rural water systems in Ghana: post-construction support and water and sanitation committees in
Brong Ahafo and Volta regions. In Water Resources Sustainability. Mays, L. W. (ed.). McGraw-Hill, pp. 267 – 288.
Lockwood, H. (2002). Institutional Support for Community-managed Rural Water Supply and Sanitation Systems in Latin
America. Dec. 2002. Strategic Report 6, Environmental Health Project Contract HRN-1-00-99-00011-00, Ofﬁce of Health,
Infectious Diseases and Nutrition, Bureau of Global Health, United States Agency for International Development,
Lockwood, H. (2003). Post-Project Sustainability: Follow-up Support to Communities. Draft. World Bank report.
D. Whittington et al. / Water Policy 11 (2009) 696–718 717
Mu, X., Whittington, D. & Briscoe, J. (1990). Modeling village water demand behavior: a discrete choice approach.
Water Resources Research,26(4), 521– 529.
Narayan-Parker, D. (1995). The Contribution of People’s Participation: Evidence from 121 Rural Water Supply Projects.
World Bank, Washington, DC. September, 108 pages.
Osafo-Yeboah, A. (1994). Using the Afridev Handpump: Norrip’s experience. 20th WEDC Conference Proceedings. Water,
Engineering and Development Centre, Loughborough University, Leicestershire, UK.
Prokopy, L. (2004). Women’s participation in rural water supply projects in India: Is it moving beyond tokenism and does it
matter? Water Policy,6, 103–116.
Prokopy, L. S., Thorsten, R., Bakalian, A. & Wakeman, W. (2008). Evaluating the role of post-construction support in
sustaining drinking water projects: evidence from Peru.Journal of Planning Education and Research,27(3), 294 – 305.
Sachs, J. (2005). The End of Poverty: Economic Possibilities for our Time. Penguin Press, New York.
Sara, J. & Katz, T. (1997). Making Rural Water Supply Sustainable: Report on the Impact of Project Rules. UNDP-World
Bank Water and Sanitation Program, Washington, DC.
Sara, J., Gross, A. & van den Berg, C. (1996). Rural Water Supply & Sanitation in Bolivia: From Pilot to National Program.
UNDP-World Bank Water and Sanitation Program.
Schumpeter, J. (1939). Business Cycles: A Theoretical, Historical and Statistical Analysis of the Capitalist Process.
Macmillan, New York.
Therkildsen, O. (1988). Watering White Elephants: Lessons from Donor-Funded Planning and Implementation of Rural Water
Supplies in Tanzania. Scandinavian Institute of African Studies, Uppsala.
Thorsten, R. (2007). Predicting Sustainable Performance and Household Satisfaction of Community-Oriented Rural Water
Supply Projects: A Quantitative Evaluation of Evidence from Ghana and Peru. PhD Dissertation, Department of City and
Regional Planning, University of North Carolina at Chapel Hill.
United Nations Development Programme (2006). Beyond Scarcity: Power, Poverty and the Global Water Crisis. Human
Development Report 2006.
United National Environment Program (2009). Sourcebook of alternative technologies for freshwater augmentation in Africa.
The International Environmental Technology Centre. http://www.unep.or.jp/ietc/publications/techpublications/techpub-8a/
wells.asp Accessed May 5, 2009.
White, G., Bradley, D. & White, A. (1972). Drawers of Water: Domestic Water Use in East Africa. University of Chicago
Whittington, D., Davis, J. & McClelland, E. (1998). Implementing a demand-driven approach to community water supply
planning: a case study of Lugazi, Uganda.Water International,23, 134– 145.
World Bank Water Demand Research Team (1993). Demand for water in rural areas: determinants and policy implications.
World Bank Research Observer,8(1), 47 – 70.
Received 11 February 2008; accepted in revised form 18 March 2008
D. Whittington et al. / Water Policy 11 (2009) 696–718718