Institutional Determinants of the Impact of Community -Based Water Services:Evidence from Sri Lanka and India

Jonathan Isham

I. Introduction

As she fills her water vessel at a standpost and then balances it carefully on her head, Arpana, a mother of five, describes how clean water arrived to her community. Three years ago, a government extension worker had informed her and her neighbors that they could get a new water service if they contributed cash or labor for its construction and took responsibility for its operation and maintenance. Arpana, with most of the members of her women’s group, attended sessions where community leaders led a discussion on the selection of the type of service and locations of the new standposts. Her husband also contributed labor to the construction of the new service. Now, she and her friends who use the water pay a monthly user fee, and a caretaker, a neighbor, looks after operation and maintenance. “My friends and I are very satisfied with this new water service, ” she notes. “It has cut the time I spend daily fetching water by over an hour, and my two youngest children no longer sufer from ‘loose stools’. “

In another community, Madhumitha is also hauling water from a standpost financed by the same government program. Madhumitha complains that, in her community, water from the standpost is not available many days because of leaks, so she and her daughter often still collect water from the local spring. She also complains about the inconvenient location of the standposts. As far as she knows, the views of community members about possible sites were never solicited. “We have a caretaker who is supposed to maintain the system “, Madhumitha says, “but he is not doing his job properly. ” But no one in this community seems to care. “Everybody just takes care of their own business and waits for others to act. Anyway, taking care of the water service is not my responsibility, ” she sighs, and slowly walks away.

Development agencies and donors currently promote a community-based approach to the provision of rural water services. This demand-responsive approach calls for a joint effort by community members and government staff in service design, construction, and operation and maintenance (O&M). Community members are typically expected to participate in the design process: in particular, to choose collectively the type and the level of service based on their willingness

to pay. In addition, communities may be asked to contribute cash or labor to construction, and take care of operation and maintenance.1 However, the outcomes of this approach have greatly varied, so development practitioners now wonder: “Under what circumstances is the community-based approach more likely to succeed?”

The objective of this paper is to start unraveling that question by analyzing selected institutional determinants of the impact and performance of communitybased water services. Using quantitative and qualitative data from 1,088 rural households and 50 water committees, the paper investigates how service rules and practices, social capital, and governmental and non-governmental organization (NGO) institutions affect the impact and performance of services supported by three World Bank-financed projects in Sri Lanka and India. 2 The paper focuses on measuring and econometrically analyzing selected aspects of project design and implementation, such as the importance of community participation in service design and decision-making.3 This paper is also one of the first to measure and econometrically analyze the effect of social capital on the impact of communitybased water services.4

II. Impact and Performance of Projects in Sri Lanka and India

A. Community-based Projects in Sri Lanka and India

In the early 1990s, three community-based rural water projects were prepared and implemented in Sri Lanka and in two states of India–Karnataka and Maharashtra. Their objectives were to provide potable water to selected small rural communities that did not have reliable access to safe water within a kilometer or

less.5

These projects adopted different ‘community-based’ strategies. The Sri Lankan households were supposed to contribute 20 percent of construction costs, either in cash or labor. The Indian households, by contrast, did not formally contribute to construction: water services were fully financed through government grant funds. In Sri Lanka and Karnataka, communities were supposed to take responsibility for O&M (including the levying of household tariffs to cover O&M costs). In Maharashtra, by contrast, district and local governments were supposed to take formal responsibility for O&M.

B. Impact and Performance

To analyze the impact and performance of water services provided by these projects, data were collected from 50 communities. Quantitative data were collected through a survey of 1088 households and 50 water committees.6 Qualitative data were gathered through focus group interviews with community members and interviews with local government officials.

Analysis of these data indicates that the impact and performance of these water services have been mixed. Means and standard deviations of several indicators of service impact and performance are reported in Table 1. Two types of self-reported household-level impact indicators are used: indicators of improved health and of time savings. ‘Improved health’ and ‘decreased incidence of diarrhea’ indicate, respectively, that the family’s health has improved and that the incidence of diarrhea has decreased since the implementation of the new water service. ‘Change of collection time’ indicates the daily change (in minutes) for

collecting water.7 Performance variables indicate project achievements at the design, construction, and O&M stages. ‘Satisfied with service design’, based on self-reported household-level responses, indicates that a household was satisfied with service design. ‘Good quality construction’, ‘no construction defects’, ‘noncolored water’, and ‘non-turbid water’, based on technical evaluations of the system by the community water committee, indicates services with these characteristics.

The results suggest that the projects in India have had a greater positive impact on health than the project in Sri Lanka. Fifty-four percent of households in Maharashtra reported that their family’s health has improved, as opposed to 45 percent in Karnataka and 36 percent in Sri Lanka. The reduction in the incidence of diarrhea was highest in Karnataka, and about the same in Sri Lanka and Maharashtra. 8

Also, projects in India have resulted in large time-savings. In Karnataka and Maharashtra, households reduced daily collection time by 62.6 and 53.9 minutes, respectively. The respective reduction in Sri Lanka was 40.6 minutes. The withinproject differences were also large (as indicated by the relatively large standard deviations). For example, in the two Indian projects, 10 percent of households report that they still spend two hours or more collecting water after the project has been implemented.

Many performance indicators , however, suggest that the performance of the Sri Lankan project has been superior. For example, 86 percent of households in Sri Lanka were satisfied with service design, as opposed to 71 percent and 45 percent,

respectively, in the Indian projects. The average quality of water delivered was highest in Sri Lanka. (although the average quality of construction was highest in Karnataka.)

How can improved health be higher among households in Maharashtra if the project has worse performance? Likewise, how can improved health and time savings be lower among the Sri Lankan households compared to the Indian households, despite generally better performance? The health impact results may be explained by lower initial health conditions in Maharashtra than in Karnataka and Sri Lanka: households that did not use a project-financed water system had significantly higher incidence of diarrhea in Maharashtra (23%) than in Karnataka (13%) and much lower incidence and medical treatment of diarrhea in Sri Lanka.9 Accordingly, the results in Table 1 are consistent with decreasing returns to health interventions: as a result of the same intervention, households with better initial health experiencing smaller health improvements than households with worse initial health. Likewise, the absolute time savings in Sri Lanka project were lower because the pre-project collection times were significantly lower than in the Indian projects: 76 minutes as opposed to 147 and 129 minutes, respectively.

C. Two case studies

The variation of performance and impacts was confirmed by qualitative beneficiary assessments. Gallella and Passaramulla, two communities served by the Sri Lankan project, provide good examples.

In Gallella, the new water service provided connections to 214 households that, prior to the project, had consumed water from unprotected wells, springs or

streams. As service design began, community members, collaborating with government and NGO representatives, agreed that household connections be provided and that the connection cost will depend on household distance from the main pipeline. Households contributed about 43 percent of total construction costs (well above the required 20 percent) in the form of unskilled labor. As the project began, the water committee in Gallella –which had many pre-existing community groups and civic activities — coordinated community participation, monitored household construction contributions, and hired caretakers to handle routine maintenance. The committee established clear procedures for tariff collection to cover O&M expenses: caretakers collect monthly fees and retain written records of payments. Ninety percent of households pay the required fee, which is the highest recovery rate among the surveyed communities, and households get together monthly to clean the water tank. Overall, water services in Gallella have had substantial impacts. Twenty-one percent of households report that the incidence of diarrhea has decreased, and the time-saving for women has been dramatic: an average daily reduction of an hour.

In Passaramulla, only one pipe-borne gravity system was in operation three years after service implementation. Seven other systems were in place but inoperable

committee but did not ensure that committee members had adequate information and training. The committee did not organize monitoring and quality control of construction, which resulted in defective construction work. The subsequent performance of the water committee has been poor. Committee members rarely meet, and financial records have been haphazardly kept. Further, operations have not been transparent: while most households do not receive any water, the committee chairman has a working household connection. The water service in Passaramulla ranks as the worst in Sri Lanka, and with very poor performance, it has had little impact.

III. Determinants of Impact and Performance: The Framework

The community-based approach to water delivery calls for collaborative design and construction among community members, government officials, and NGO staff. Their incentives will determine whether, in practice, they actually collaborate, and institutions affect these incentives. In the delivery of communitybased water services, institutions are the formal and informal rules and practices that govern behavior of different groups.10 By limiting opportunistic behavior, they can hold the groups to their commitments in the design, construction, and O&M of water services, thereby improving service performance and impact.

Figure 1 illustrates a chain of causality from three sets of institutions–service rules and practices, social capital, and governmental and NGO rules and practices — to service performance and impact. Each box in Figure 1 contains a list of indicators and proxies used in this paper to measure these determinants (as well as non-institutional determinants.)

First, water service performance will impact household health and timesaving (link 1). A well-designed, well-constructed, and well-maintained water service that is conveniently located for most households and provides a constant flow of clean water is likely to improve household health and reduce water collection time.

Second, service rules and practices water will influence service performance (link 2). In community-based water services, users, in collaboration with government officials and NGO staff, are expected to craft rules and practices about user participation in decision-making, design, construction, and O&M

Third, social capital is likely to influence the existence and effectiveness of service rules and practices (link 3). Social capital refers to the norms and networks that facilitate collective action.13 Community-level social capital is likely to help community members to craft and enforce the service rules that govern the design, construction, and O&M. The collective demand for the type and level of services is more likely to be clearly expressed when community members are accustomed to working together, where leaders are accountable, and where all stakeholders have a voice. Water users groups are more likely to succeed in communities with cohesive community groups and regular civic activities. Formal and informal social ties deter

community members from free riding and constrain community leaders from shirking and expropriating funds.

Fourth, governmental and NGO institutions are also likely to affect the existence and effectiveness of service rules (link 4). Government officials and NGO staff helped to implement these projects: facilitating the establishment of a functioning water committee

Finally, non-institutional determinants are also likely to affect the existence and effectiveness of service rules (link 5). 14 These include household assets, household size, level of human capital, and environmental conditions. For example, the availability of alternative water source will affect a community’s willingness to craft effective system rules and practices.15

IV. Determinants of Performance and Impact: Empirical Evidence

Does the framework presented in the previous section hold in practice? This section provides empirical evidence for the linkages in the framework, using data from the household and water committee surveys.16

A. Proximate determinants of impact

The first link in the framework ties service performance to impact. To estimate the proximate determinants of health impacts, begin with an econometric

model based on the following relationship:

Hij* = 0 + Dj 1 + Cj 2 + Xij 3 + ij, (1)

where Hij* is a latent random variable for household i in community j which is some measure of the changed health of the household since the implementation of a community-based water service. Assume that Hij* is a linear function of a set of non-stochastic independent variables and an error term (ij). These covariates include (as discussed in the previous sections): Dj, design performance of the water service in community j

The dichotomous variable ‘improved health,’ is used as the dependent variable (with Probit estimation) to test the relationship presented in equation (1), because the available data do not include continuous measures of the change of household health.17 The community-level independent variables used to test these relationships (summarized in Appendix Table 1) are ‘community design satisfaction’, the share of households in each community that were satisfied with project design

The results of testing the linkage between performance and health impacts (equation 1), are listed in Table 2 and summarized as follows: 19 Improving community satisfaction with service design enhances the service’s

health impact. ‘Community design satisfaction’ is a significant and positive determinant of improved health in all three projects. Based on the standard deviations reported in Appendix Table 1 and the change in probabilities reported in Table 2, a one-standard deviation increase in ‘community design satisfaction’ is associated with an increase in the probability of improved health of 0.09 in Sri Lanka, 0.13 in Karnataka, and 0.11 in Maharashtra..20

§ Ensuring that water services are well constructed enhances the service’s health impact. ‘Good quality construction’ is a significant and positive determinant of improved health in Sri Lanka and Maharashtra (and positive in Karnataka).21 A change from bad quality (the presence of serious construction defects) to good quality (the absence of serious construction defects) construction is associated with an increase in the probability of improved health of 0.13 in Sri Lanka and 0.18 in Maharashtra.22

§ Providing hygiene training (or ensuring that hygiene training is provided by other sources) enhances the service’s health impact. Enrollment in a hygiene class is associated with an increase in the probability of improved health of 0.13 in Sri Lanka and 0.20 in Maharashtra.23

§ Non-institutional household variables (household size and assets) are not significant determinants of improved health in any of the three projects. This is true also of indicators (not reported here) such as household demographics and wealth24 and the type of previous drinking water source used by the household (for example, hand-dug well or spring).25

A similar econometric framework is adopted to estimate the proximate

determinants of time-saving impacts. In this case, the econometric model is based on the following relationship:

Tij = 0 + Dj 1 + Xij 3 + ij, (2)

where Tij is a continuous measure of the time-saving of household i in community j.26 The estimation procedure must account for the fact that time savings are likely to be greater in households in which the pre-project collection times are significantly higher, as discussed before. Accordingly, using the logarithm of time-saving as the measure of Tij allows one to estimate the percentage change of time-saving per household.27

The results of testing the linkage between performance and time savings (equation 2) are listed in Table 3 and summarized as follows:

Improving community satisfaction with service design reduces water collection times. ‘Community design satisfaction’ is a significant and positive determinant of time-saving in all three projects. A one-standard deviation increase in ‘community design satisfaction’ is associated with a decrease of collection time of 19 percent in Sri Lanka, 45 percent in Karnataka, and 32 percent in Maharashtra. Based on the means of the pre-project collection times, households, on average, will save 15, 67, and 41 minutes, respectively, with such an increase.

The determinants of household time savings – based on a community-level decision about the placement of a new water system – are not at the household level. With the exception of ‘household size’ in Karnataka and ‘hygiene class’ in Maharashtra28, household variables are not significant determinants of time

saving. Again, this is true of the variables reported here, as well as alternative household indicators (not reported here).

Overall, the results in this section suggest that well-designed and wellconstructed water services are likely to improve household health, and that welldesigned water services are likely to lower collection times. They also underline the importance of providing hygiene classes in conjunction with a water project for improving household health. While these conclusions are certainly not ground breaking, these results allow one to establish the statistical significance and relative magnitudes of the importance of well-designed and well-constructed water services across three different projects.29 More importantly, these results allow one to test econometrically the less explored linkages of this framework: how institutions underlie the performance indicators.

B. Institutional determinants of performance

Does community participation and decision-making lead to higher satisfaction with service design, as suggested by the framework? To answer this question, three household-level dummy variables were created from survey questions about the service design process. First, ‘local initiation’ indicates that community members, as opposed to government officials or other outsiders, had the original idea to build the water system. Second, ‘design participation’ indicates that the household participated in service design. Third, ‘local decisionmaking’ reflects that community members, as opposed to government officials or other outsiders, made the final decision about what type of system to build.30

Table 4 reports results from probit estimates of the household-level

determinants of ‘satisfaction with service design’, with community fixed effects. The results can be summarized as follows31:

§ Households are no more satisfied with service design when the original idea to build a system comes from community leaders rather than from outsiders.

§ User participation in design leads to greater satisfaction with service design. A discrete change from not participating to participating leads to an increase in the probability of being satisfied with service design of 0.196, 0.253, and 0.419 in Sri Lanka, Karnataka, and Maharashtra, respectively.32

§ Letting locals make the decision about the system type leads to greater satisfaction with service design. A discrete change from stating that local decision-making did not prevail to stating that it did leads to an increase in the respective probabilities of 0.191, 0.322, and 0.540.

These results conform to the analytical framework. Households are more likely to be satisfied with service design when they have participated in the design process and when the community makes the final decision about service type. This is true within each project and within each community (given the use of community fixed effects), despite different approaches to service design among the projects. In addition, these results indicate that the initiation of well-designed services can begin from outside or inside of the community, as long as local participation in design and decision-making is ensured.

What are the institutional determinants of good construction? Is construction better when household contributions are monitored and sanctions against misconduct are imposed, as suggested by the framework? Since ‘good

quality construction’ is a community-level variable, the sample size for addressing these questions econometrically must be 50, the number of communities in the sample. Table 5 lists the within-project associations between ‘good quality construction’ and two indicators of service rules and practices. ‘Construction monitoring’ is the community share of households that said that the required construction contributions (cash or labor) were monitored by other community members. ‘Construction sanctions’ is the community share of households that said that households that did not contribute their share were charged a financial penalty.33

The analysis yields the following results about the determinants of ‘good quality construction’:

§ Existence of monitoring mechanisms leads to better quality construction. A one-standard deviation increase in ‘construction monitoring’ increases the probability of ‘good quality construction’ by 0.38.

§ Existence of construction sanctions does not measurably improve construction quality.34

Overall, the results in this sub-section show that community participation and decision-making in service design lead to well-designed services, and monitoring of household contributions to construction lead to better-constructed services.

C. Social capital and service rules

Finally, the framework suggests that existence of service rules depends on social capital. This section tests if social capital is a significant determinant of

‘design participation’ and ‘construction monitoring’.

An econometric model based on the following relationship is used to assess the influence of social capital on service rules:

Pij* = 0 + Sij 1 + Xij 2 + Xj 3 + ij ,

(3)

where Pij* is a latent random variable for of household i in community j which is some measure of the intensity35 of design participation

The primary indicator of social capital used is the ‘social capital index’, a composite index of the quantity and quality of local groups (based on the ‘Putnam index’ in Narayan and Pritchett 1999), that attempts to capture the underlying behavior of interest: that a household has established a pattern of working cooperatively with other households and community leaders. As summarized in the second part of Appendix Table 1, this indicator is created as follows. First, ‘number of groups’ is the number of community groups to which a household belongs. This includes economic groups (such as, farmer’s groups and credit/finance groups), religious groups, and social groups (such as, women’s

groups and youth groups). Second, ‘group characteristics’ is an additive subindex of various characteristics of each household’s most important group, including heterogeneity of members by caste and religion, heterogeneity of members by occupation, the nature of decision-making mechanisms, and effectiveness of group functioning. The additive sub-index is increased by one unit if a household’s most important group has: caste groups that are proportionally represented

Summary statistics for ‘number of groups’, ‘group characteristics’, and the social capital index reveal a dramatic difference in the quantity of associational activity in Sri Lanka and India (see Appendix Table 1). On average, households in Sri Lanka belong to 2.4 groups. In Karnataka and Maharashtra, this figure is 0.19 and 0.49, respectively. The means of group characteristics and the social capital index are: 7.48 and 25.38

An alternative social capital indicator is ‘help from outsiders’, a dummy

variable that indicates that a household could get help from non-family members

in difficult times. Community members that can do so are likely to have

established productive norms and networks with neighboring households. The

project-level means for this indicator are 0.61, 0.62, and 0.60, respectively. The results summarized in Table 6 reveal that social capital and design participation are associated. Higher household-level social capital is positively associated with participation in the service design. Specifications

(1), (3) and (5) shows a statistically significant relationship between the ‘social capital index’ and ‘design participation’. A one-standard deviation increase in the ‘social capital index is associated with increases of 0.06, 0.08 and 0.13, respectively, in the probability of design participation (compared to project means for design participation of 0.84, 0.11 and 0.21).

The statistically significant relationship between social capital and design participation survives the inclusion of other potential covariates. Specifications

(2), (4) and (6) reveal that the inclusion of ‘household assets’ and ‘family size’, with community fixed effects, does not change the basic relationship between the ‘social capital index’ and ‘design participation’.40, 41

The robustness of these results is confirmed in two ways. First, in all six specifications, replacing the ‘social capital index’ with either of its the sub-indices or ‘help from outsiders’ yields the same statistically significant relationship between a measure of social capital and design participation.42 Second, in the two specifications for Sri Lanka, the only project that required household participation

in construction, replacing the ‘design participation’ with the equivalent ‘construction participation’ yields a statistically significant relationship (not reported here). A one-standard deviation increase in the social capital index is associated with a 0.09 increase in the probability of construction participation. Two of the three alternative social capital indicators (‘number of groups’ and ‘help from outsiders’) also yield statistically significant relationships.

Also, community-level social capital is a positive and significant determinant of construction monitoring. Table 8 lists results from communitylevel specifications–in India and Sri Lanka, respectively–of the determinants of construction monitoring: in addition to the community-level social capital indicator, each specification includes (not shown) community-level averages of assets, household size and (in the case of India), a dummy variable for Karnataka. With two of the four indicators in Sri Lanka and three of the four indicators in India, community-level social capital is a positive and significant determinant of construction monitoring.

The results in this sub-section show that household-level social capital leads to participation in service design: in communities with effective community groups, participation in service design is likely to be higher. The results from Sri Lanka show –that social capital also tends to increase participation in construction design. Finally, social capital is positively associated with construction monitoring.

D. Magnitudes of the effect of institutions on impact

The previous results suggest a chain of causality from institutions –(social

capital and service rules) to project performance and impact. This section calculates the implied magnitudes of the effect of institutions on service impact, improvement of household health and reduction of water collection time.

The first part of Table 8 presents some calculations of such magnitudes for improved health. The first row of the table presents the effect of a one-standard deviation increase in design participation on improved health, based on the underlying structural equations summarized in Tables 2 and 3. Note that the three project figures, from 0.051 to 0.068, are within a plausible and fairly narrow range, despite the large difference in the nature of the quantity and quality of the underlying variables across these three projects (particularly between Sri Lanka and the two Indian projects). For each project, this is calculated by multiplying the standard deviation of design participation with the coefficients within the framework that lead from ‘design participation’ to ‘design satisfaction’ (link 2) and from ‘design satisfaction’ to ‘improved health’ (link 1). For example, in Sri Lanka, the figure is 0.051 = (.360)*(.196)*(.72), where the three respective multiplicands are: the standard deviation of design participation

health for 17 to 18 households in a community of 200 households (the means community size in the sample) because of more design participation and better construction monitoring.

The final two rows of Table 8 present similar calculations of these magnitudes for time-saving, based on similar underlying equations. The calculations from the structural equations, from 0.012 to 0.024, are also within a plausible range, despite the large differences among the projects. These magnitudes translate to time-saving of 9, 35, and 25 minutes, respectively, from more community-level design participation.

Since both ‘design participation’ and ‘construction monitoring’ are endogenous in this framework (that is, they are determined by social capital and other factors), another way to calculate the effect of institutions on service impact is with ‘reduced form’43 estimations, where the community-level social capital index replaces both ‘community design satisfaction’ and ‘good quality construction’ in the specification tested in Table 2 and 3. Single-stage reduced form estimates are consistent on the condition that: the indicators of social capital accurately measure the patterns of social interaction and norms of trust and reciprocity among water users and their neighbors

Table 9 presents the estimates of single- and two-stage reduced form models. In single-stage estimation, probit models show that only in the case of Maharashtra is the village-level social capital index positive and significant

(coefficients of 0.0269 and 0.0244 respectively).45

In the IV estimation, two instruments for the social capital index are used, based on additional survey questions about community activities. ‘Household community activity’ is the community-level average of households that participated in a community-level activity

Table 9 shows that the IV results are positive and significant in the case of Sri Lanka and Maharashtra, with coefficients of 0.0286 and 0.0417, respectively.47 Using these IV results, a one-standard deviation increase of community-level social capital is associated with an increase 0.17 and 0.13 in the probability of improved health. 48 These magnitudes are slightly larger than the sum of the magnitudes of design participation and construction from the structural equations.49 Accordingly, the evidence from the two-stage reduced form equations, in two of the three projects, suggests that more social capital — the critical determinant of design participation and construction monitoring — leads to improved household health for about 26 to 34 households in a community of 200 households.

The final two rows of Table 9 present calculations from comparable OLS and IV estimates for time-saving. In single-stage estimation, the village-level

social capital index positive and significant in the cases of Sri Lanka and Maharashtra (coefficients of 0.0047 and 0.0916, respectively). In the IV estimation, this is true of Maharashtra (0.0807). Using the IV result, a onestandard deviation increase of community-level social capital is associated with a reduction of water collection time by 26 percent. This corresponds to daily timesaving of about 33 minutes.

V. Conclusion

Using data from Sri Lanka and India, this paper has shown that welldesigned and well-constructed water services lead to improved household health and reduced water collection times. The results suggest that one can promote well-designed services—(that is, increase user satisfaction with the service design) by involving community members in the design process and by letting community members, not outsiders, make the final decision about the service type. Ensuring that communities have effective mechanisms to monitor household contributions to construction is in turn an effective way to promote well-constructed services.

However, household participation in service design and ability to craft and enforce monitoring mechanisms are not automatic. The empirical results presented here suggest that in communities with high levels of social capital–in particular, with active community groups and associations–design participation is more likely to be high and monitoring mechanisms are more likely to be in place. In those communities, households are accustomed to working together and social ties deter free riding. This suggests a way to place an economic value on

community-level social capital in the context of water projects: as the net present value of the marginal increase in health associated with active civic associations.50

What do these results, in particular the results about social capital, imply for designers of community-based water projects? They do not necessarily suggest that projects should avoid investing in community-based water systems in communities with low levels of social capital. Indeed, while many poor communities with the most urgent need for improved water systems are likely to have low levels of social capital51, people in many of these communities are likely to reliably report a willingness to pay and maintain a water system. Instead, these results suggest that designers of community-based water projects need to pay attention to the prevailing levels of social capital, as one of the factors that will influence the performance of the project, in communities to be served by the project. When targeting these communities, the allocation of investment resources for water services programs may need to be adjusted to take into account the lack of social capital. Possible adjustments include increased investments in social mobilization efforts (for example, through the strengthening of local organizations) and in more direct supervision by project personnel working in these communities to oversee system performance.52