2.1. Ethics statement

The Institutional Review Board for the Council of Scientific and Industrial Research in Ghana reviewed and approved our research protocol (RPN 022/CSIR-IRB/2022). Prior to any field work, the study was introduced to District Assembly members, and field work supervisors described the work and requested permission from community leaders. We obtained written informed consent from all survey participants, and all collected data was kept confidential. We have shared initial results with local authorities in aggregated form, with identifiable information removed. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S1 Checklist).

2.2. Study location

This study primarily uses cross-sectional household surveys conducted during baseline data collection for a program focused on improving water and sanitation in rural northern Ghana. This baseline effort occurred in the Upper West, Upper East, Northern, and North East regions of Ghana, with surveys taking place in two districts within each region. These districts included: Wa East and Daffiama Bussie Issa from the Upper West Region, Bawku West and Tempane from the Upper East Region, Mion and Yendi from the Northern Region, and East Mamprusi and Mamprugu Moagduri from the North East Region. In each district, we began with a list of rural communities which were planned to be targeted for program activities in the near future due to poor existing water and sanitation conditions (e.g., high open defecation rates), and we randomly selected 14–15 communities to include in the survey. One district in the North East Region (East Mamprusi) contained only 14 communities targeted for intervention, while all other districts included 15 communities.

In total, we surveyed 120 communities in northern Ghana. Because communities were selected based on planned program implementation, they are not necessarily representative of the study districts or regions as a whole. Instead, they represent communities with some of the poorest levels of water and sanitation access, which district officials identified as most in need of improvements. Broadly, all selected communities were rural, and most were in relatively remote locations, although a small number of rural small towns were included. Household surveys occurred from December 2022 to February 2023, during northern Ghana’s dry season, when water availability may be most challenging.

In addition to the household surveys (described in further detail below), we also conducted focus group discussions and follow-up interviews with members of the District Interagency Coordination Committee on Sanitation (DICCS). We integrated information from these activities with our household results by creating binary district-level variables focused on topics such as WASH-related policies, budgets, and monitoring.

2.3. Household and respondent selection

Within each community, we surveyed 13–17 households, for a total of 1,748 households across the entire study area. We randomly selected households according to the following procedure: At the beginning of a new survey, the CommCare survey application (used to conduct the surveys on Android phones) assigned each enumerator a randomly-generated direction (out of eight possible cardinal directions, including north, northeast, east, etc.) and a random number of households. This number was based on the community size: In communities with less than 20 households, this number could range from 1 to 10, and in communities with 20 or more households, it could range from 1 to 25. The enumerator then walked in the assigned cardinal direction until they passed the assigned number of households, and the last household they counted was asked to participate. This procedure for household selection was similar to random walk procedures others have used in similar rural settings to generate spatially random results comparable to population representative values [12], with some modifications to our local context. Community guides informed enumerators of the community boundaries before surveys began, and communities were generally open or had many dirt paths between households to allow enumerators to move freely throughout the community in each randomly generated direction without needing to follow specific routes or roads. In communities with less than 13 households, we surveyed all available households.

After selecting a household, the enumerator aimed to target an adult female involved in the household’s financial decisions. We targeted female respondents because women in this context are often most involved in household water and sanitation provision. If such an individual was not available, the enumerator surveyed another adult household member who was at least 18 years of age and was knowledgeable about financial decisions. Prior to beginning the survey, the enumerator explained the study, confirmed the participant’s age, and obtained written informed consent.

2.4. Survey design

The baseline survey included numerous questions concerning water, sanitation, and hygiene. For the present study, we asked about households’ current water supply characteristics, practices, and payments, as well as levels of satisfaction with these existing services and willingness-to-pay (WTP) for hypothetical alternative services. To assess satisfaction, we asked about respondents’ overall levels of satisfaction with respect to their current water service using a four-point Likert scale (very unsatisfied, somewhat unsatisfied, satisfied, very satisfied). We also asked follow-up questions regarding which aspects of the service (accessibility, availability, reliability, safety, and affordability) they liked or disliked.

To estimate WTP for water services with various hypothetical characteristics, we used discrete choice experiments (DCE, described in detail in the following section) among a random subset of baseline survey respondents (N = 542, 31% of all baseline respondents), as this portion of the survey required a considerable amount of time to complete. In addition to the DCE implementation within the baseline survey, we also conducted 424 supplemental DCE surveys in northern Ghana focused specifically on small town residents with greater exposure to piped water services. These additional surveys enabled us to investigate whether this slightly different context would alter how individuals valued the water service features included in the hypothetical DCE scenarios. The supplemental DCE surveys followed a similar approach for random household selection with modifications to ensure that both on-premises and off-premises water users were captured. Unlike the baseline sample that targeted high-need pre-intervention communities, the supplemental small-town DCE communities were randomly selected from a master list of towns with water systems in the study area to represent different water system types, sizes, and suppliers. These surveys took place in May-June 2023 in the Upper West, Upper East, North East, and Savannah regions.

In addition to household surveys, enumerators also observed community water points. For each community water point, enumerators completed a separate CommCare form focused on water point infrastructure, functionality, and upkeep. Key water point observations included whether water points were functional, well-maintained, and clean at the time of data collection, and whether they were easily accessible (e.g., not obstructed by vegetation or debris).

Across all data collection for this study, participant recruitment began on December 21, 2022 and ended on June 20, 2023.

2.5. Discrete choice experiment design and implementation

The DCE method enabled us to consider the impact of separate features reflecting different types of improvements in moving toward safely managed water services (Table 1). In this approach, the enumerator presented a survey respondent with a series of hypothetical choices between two scenarios. Each scenario described a combination of five service features (Table 1; Table A in S1 Appendix). The two scenarios presented to the respondent are known together as a choice set (see Figure A in S1 Appendix for an example).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Water service features considered in discrete choice experiments to estimate willingness-to-pay. Prices are shown in Ghanaian Cedis (GHS; 11 GHS = 1 USD in 2023 during data collection).

https://doi.org/10.1371/journal.pwat.0000547.t001

A full factorial design, that would include all possible combinations of the features, would have required 192 distinct scenarios and over 18,000 unique choice sets. Instead, we employed a Fedorov exchange algorithm [13] to estimate an optimal partial design that considered: (i) orthogonality, minimizing the possible correlation between features, and (ii) level balancing, ensuring that the levels within each feature appeared a similar number of times. We ran the algorithm multiple times to assess possible partial designs containing up to 30 scenarios, repeating the procedure up to 1,000 times at each potential sample size to ensure it arrived at the best possible outcomes. The selected design contained 24 scenarios (Table A in S1 Appendix), resulting in a total of 276 unique choice sets. Finally, we removed 35 choice sets where one scenario was superior or equivalent to the other with respect to every feature (including price), as we would always expect respondents to choose the superior scenario (providing no useful information for understanding tradeoffs). This process left us with a total of 241 choice sets included in the survey.

After walking a respondent through a simple introductory example (involving a choice of two meals), the enumerator then presented a series of four choice sets, each randomly selected by the CommCare application. A set of 24 informational handouts aided the enumerator in describing the two scenarios within each choice set (examples provided in Figure A in S1 Appendix). Following the explanation of a choice set, the respondent stated which scenario they would prefer. In our survey, they were also able to opt out of the choice, meaning they would not pay anything but not gain access to either scenario (in this hypothetical situation). After completing the four choice sets, the enumerator repeated the first choice set as a quality assurance check, confirming that the participant responded in the same way. Only 3% of respondents were inconsistent in choice.

In addition to the DCE choice sets, we also asked open-ended questions about WTP for two specific scenarios from our set of 24. These two scenarios represented the highest and lowest overall levels of service:

• Highest level of service (Scenario 1 in Table A in S1 Appendix): What is the maximum amount that your household would be willing to pay for access to chlorinated water from a tap inside your dwelling, where the water service is available 24 hours per day and is interrupted no more than 1–2 days per month, and breakdowns never last more than 24 hours per day because someone performs regular maintenance?
• Lowest level of service (Scenario 6 in Table A in S1 Appendix): What is the maximum amount that your household would be willing to pay for access to unchlorinated water from a public handpump, where it takes 30 minutes to go, collect water, and return home; the water is typically available for 4 scheduled hours per day; water service is interrupted 2 days per week and breakdowns can last more than 24 hours because maintenance only happens when needed?

WTP for the highest level of service was used as a consistency check for DCE responses. If the respondent had selected DCE scenarios with higher price points than their WTP for the highest level of service, the enumerator went back and confirmed those answers. After going back, respondents confirmed their selection of higher price points than their maximum reported WTP in 10% of choice sets. We removed these discrepancies during analysis. WTP for the lowest level of service was used as a reference level, allowing us to add on WTP increases for specific service features derived from the DCE analysis.

2.6. Data analysis

3.1. Characteristics of household respondents

Across our study population, 77% of 1,748 total baseline survey respondents were female, while only 13% were members of female-headed households (Table 2; these results focus only on baseline respondents and do not include supplemental DCE participants; characteristics of supplemental participants are provided in Table C in S1 Appendix). According to our computed wealth quintiles, Upper East households tended to be less wealthy than those in other regions. Across all regions, household heads were typically employed in agriculture and had completed no formal education. Local officials classified most communities as remote rural, but 23% of respondents lived in communities classified as rural small towns, which tended to be larger and located along main roads.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Descriptive statistics showing baseline survey respondent characteristics across the entire study population and within each of the four regions. To compare characteristics across regions, we applied chi-squared tests in R, with p-values reported in the final column. We used UN JMP definitions to define basic and safely managed water services. Because we did not have data on water source contamination, the table reports the maximum possible percentage using safely managed services.

https://doi.org/10.1371/journal.pwat.0000547.t002

Most households (62%), especially those in the Upper West and Upper East regions, reported using a borehole with a handpump as their primary water source (Table 2). Surface water and unprotected dug wells, both of which are classified as unimproved sources, were the second (16%) and third (11%) most common sources, respectively. The Northern region exhibited the highest percentages of surface water use (38%), while the greatest use of unprotected dug wells occurred in the North East (29%). Overall, piped water was uncommon (<4%), with only 4 respondents having access to piped water on premises (on plot or in home). However, nearly 10% of North East respondents reported using piped water from public taps. More than half of respondents reported paying for water services, although most of these (87%) only paid when repairs were needed.

We classified approximately half of respondents (52%) as having access to at least basic water services (Table 2). Most respondents used improved sources such as boreholes, and a majority (70%) reported collection times within 30 minutes, which is also necessary for basic access. Collection times tended to be longer in the Upper West region (only 50% within 30 minutes). Across the study area, respondents who used surface water, unprotected springs, tanker trucks, or water kiosks also tended to have longer collection times (50% within 30 minutes for this group of sources, compared with 74% for others). In 93% of households, adult women were largely responsible for the task of collecting water.

We classified very few households (<2%) as having access to safely managed water services, primarily because only 2% met the safely managed criterion of accessibility on premises (Table 2). The availability criterion was less limiting, with most respondents (72%) reporting that sufficient drinking water was always available when needed. Our results do not provide direct evidence of water quality and the third safely managed criterion (free from contamination). However, only 8% of respondents reported treating their water.

Despite having water services that rarely met more than one safely managed criteria, approximately half of household respondents (51%) reported being satisfied or very satisfied with their current services, although satisfaction levels were considerably lower in the Upper West region (Table 2). In the following sections, we explore the factors that were associated with satisfaction, as well as the relative value respondents placed on these factors, through analysis of (i) existing water services and satisfaction levels, and (ii) hypothetical scenarios with varying service features. Insights gained through these analyses could provide valuable information regarding how to progress toward safely managed services along a path that is most aligned with user priorities.

3.2. Factors associated with existing satisfaction levels

An initial comparison of satisfaction levels and water source showed that households tended to be more satisfied when they had access to an improved water source (Fig 1). Satisfaction was low among those using surface water (26% satisfied or very satisfied), as well as unprotected dug wells (32%) and unprotected springs (0%). In contrast, satisfaction was particularly high among those with piped water access (81% satisfied or very satisfied), even when off premises (79%). Satisfaction was also high among those using boreholes (61%), especially for the 4% with access to mechanized boreholes (i.e., where the person fetching does not have to pump; 81% satisfied or very satisfied).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Level of satisfaction with water services for household respondents using different types of water sources for drinking.

To more easily visualize the relative number of respondents using each source type as well as the satisfaction levels associated with each source, the graphs provide both the number and percentage of respondents in each category.

https://doi.org/10.1371/journal.pwat.0000547.g001

Multivariate logistic regression confirmed these trends when accounting for other variables (Fig 2). Unimproved water sources such as unprotected dug wells and surface water were associated with lower odds of satisfaction than boreholes with handpumps (the most common source type) and other improved sources. For example, using surface water reduced the odds of being satisfied by 94%, compared with hand-powered boreholes (odds ratio [OR]: 0.062, 95% CI: 0.033-0.116, p < 0.001). Unprotected dug wells had a similar effect (OR: 0.146, 95% CI: 0.089-0.242, p < 0.001), while protected dug wells were also associated with somewhat lower satisfaction than other improved sources (OR: 0.431, 95% CI: 0.207-0.899, p = 0.025). Notably, we were unable to include certain source types, including piped water on premises, unprotected springs, and sachet water, in this regression model, because they were used by very few respondents.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Variables associated with water service satisfaction among 1,748 households, based on logistic regression.

For the purpose of this regression, the four-level Likert scale was consolidated into two levels: satisfied (including “satisfied” and “very satisfied”) and unsatisfied (including “somewhat unsatisfied” and “very unsatisfied”). All variables shown were included in the same multivariate regression model, which included adjusted standard errors to account for community clustering. Additional variables were considered in earlier models (Table B in S1 Appendix), and variables with limited association with satisfaction were removed sequentially until the model only contained variables with p < 0.20. Variables exhibiting high levels of collinearity (variance inflation factor >3) were also removed. Variables with odds ratios above 1 are positively associated with satisfaction, while odds ratios below 1 are negatively associated. Error bars represent 95% confidence intervals. Reference levels for categorical variables with more than two possible values (region, water source) are defined to the right of the graph. Variables are grouped by scale (region, household, community).

https://doi.org/10.1371/journal.pwat.0000547.g002

Beyond the type of water source, satisfaction levels were also associated with several other factors, some of which corresponded to criteria for safely managed or basic access. Satisfaction levels tended to be higher among households who met the availability criterion, reporting that they always had drinking water available in sufficient quantities (OR: 2.66, 95% CI: 1.94-3.66, p < 0.001). The degree of water source accessibility also seemed to be an important consideration for our survey population. Respondents able to collect water within 30 minutes had over three times the odds of being satisfied, compared with those who were further away (OR: 3.09, 95% CI: 2.28-4.19, p < 0.001). Additionally, those few with access to water on premises had higher satisfaction, although the small sample size limited our ability to precisely estimate the associated odds ratio (OR: 2.90, 95% CI: 1.26-6.66, p = 0.012).

With respect to household characteristics, satisfaction levels tended to be somewhat lower among female respondents (OR: 0.760, 95% CI: 0.578-0.999, p = 0.049). Households that reported paying for water also tended to be less satisfied (OR: 0.574, 95% CI: 0.393-0.838, p = 0.004). However, of the households paying for water, most (87%) reported only doing so when repairs were needed.

Additionally, those living in communities where a higher fraction of households paid for water also may have had somewhat higher satisfaction (OR: 1.77, 95% CI: 0.975-3.21, p = 0.061). Interestingly, though, a respondent’s satisfaction tended to be lower if they lived in a wealthier community, where a greater fraction of households fell in the top wealth quintile relative to the rest of the study population (OR: 0.480, 95% CI: 0.249-0.926, p = 0.028). As small towns tended to be wealthier than more remote communities (p < 0.001), it is possible that small town residents aspired to more “urban” water systems, such as a piped supply distributed to on-premises connections. The small town variable itself was not included in the final model, because it was highly correlated with other variables such as community wealth (p < 0.001), and it was not associated with satisfaction when controlling for these other variables. At a regional level, satisfaction levels in the Upper West and Upper East regions, and to a lesser extent in the North-East region, were considerably lower than in the Northern region.

Finally, we were unable to include local types of water service provider directly in our regression model, because this variable was highly correlated with region. However, we performed a separate regression focused solely on this variable’s association with satisfaction (Figure B in S1 Appendix). Relative to private and non-governmental safe water enterprises (the most common provider type among our respondents), community Water and Sanitation Management Teams (WSMTs) were associated with lower satisfaction (OR: 0.346, 95% CI: 0.238-0.484, p < 0.001). In Ghana, WSMTs are composed of community members and often have direct responsibility for maintaining rural community water systems, with oversight from District Assemblies [6]. Nearly all households served by WSMTs (99%) were located in the Upper West region, likely contributing to the lower satisfaction levels observed there. Water systems directly managed by District Assemblies had higher satisfaction levels (OR: 1.95, 95% CI: 1.17-3.23, p = 0.010). District-managed systems were located primarily in the Upper East (68%), although 30% were in the Northern region. Finally, and perhaps unsurprisingly, satisfaction levels were lower when water systems had no defined manager (OR: 0.331, 95% CI: 0.141-0.774, p = 0.011). The Upper West region contained approximately half of these systems.

3.3. Preferences concerning existing aspects of water service quality

As a complement to the regression analysis, we also analyzed how satisfaction levels related to direct statements made by respondents regarding what they liked and disliked about their existing water services. Immediately after respondents stated their satisfaction level in the survey, they were asked to select the aspects of their water service they liked, and then the aspects they disliked. The aspects they could select included accessibility, affordability, availability, safety, and reliability, and they had the option of selecting no or multiple items.

Accessibility and safety appeared to be most closely related to overall satisfaction levels. When people were unsatisfied, we saw frequent mention of disliking these service aspects, while satisfied respondents often reported liking these aspects about their water service (Fig 3). For example, 51% of satisfied respondents reported liking their water service’s accessibility, and 68% liked its safety. Conversely, 60% of unsatisfied respondents disliked accessibility characteristics of their water service, while 56% disliked its perceived safety (or lack thereof). In other words, these aspects seemed to be most important to people and may have had the most influence over whether or not they felt satisfied with their current service. It is important to note that respondents were focused on accessibility in terms of how easily the source could be reached and the absence of obstructions, rather than whether or not it was located on premises.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Relationships between household satisfaction with water services and the aspects of water service quality that households mentioned liking or disliking about their current service.

Each plot shows how many households within each satisfaction level reported liking or disliking a specific aspect about their water service. For simplicity, we consolidated satisfaction into two levels: satisfied (including “satisfied” and “very satisfied”) and unsatisfied (including “somewhat unsatisfied” and “very unsatisfied”). Accessibility and safety were most closely related to overall satisfaction levels, followed by reliability and availability.

https://doi.org/10.1371/journal.pwat.0000547.g003

We saw similar trends with availability and reliability (Fig 3). However, the large proportions of respondents who did not mention liking or disliking these aspects suggested that they were less critical, having a smaller effect on satisfaction levels. The results also suggested few affordability concerns, with very few people mentioning disliking their water service’s level of affordability.

3.4. Preferences and willingness-to-pay for hypothetical water services

Beyond considering existing conditions, we also used DCE methods to explore respondents’ preferences around hypothetical changes to water services. In this section, we focus first on DCE results from our baseline survey (N = 542), and then we discuss results from additional DCE surveys in rural small towns with greater levels of piped access (N = 424). In many cases, we discuss our results in relation to the reference scenario (consisting of a public, unchlorinated handpump requiring a 30-minute collection time, where water is typically available for 4 set hours per day, but unscheduled service interruptions occur 2 days per week and can last more than 24 hours), which represented the lowest level of service among our 24 DCE scenarios. The features of our reference scenario also aligned well with typical conditions among our baseline survey population, such as the common use of handpumps with collection times of approximately 30 minutes. Mean stated WTP for this reference scenario was 8 GHS/m³ among baseline participants.

Broadly, despite high levels of poverty in our baseline study area, model results suggested that people would be willing to pay more for improved service quality (Fig 4). However, it is important to remember that the results presented here reflect participants’ stated preferences when no real money was involved. Previous research has found that these types of methods can overestimate respondents’ revealed or actual WTP [19]. While we report estimated WTP values below, we primarily use these to inform our understanding of how respondents prioritize different service features relative to one another, rather than as recommendations for tariff increases associated with such improvements.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Estimated willingness-to-pay for service improvements relative to a baseline level of service (defined as access to an unchlorinated public handpump that is a 30-minute roundtrip from the household, with water available 4 hours per day and breakdowns 30% of the time), based on discrete choice experiments among baseline survey participants (N = 542).

The dashed red line shows mean WTP for the reference scenario, as stated by respondents in response to an open-ended question. Points represent estimates of mean WTP when an improved feature is added, and error bars represent the first and third quartiles of the modeled WTP distribution (modeled as a normal distribution based on individual preferences). Accordingly, results where error bars cross the baseline mean (e.g., 95% reliability) suggest that at least 25% of respondents may not be willing to pay more for these features, even if a majority do value them more highly than the baseline level. Note that error bars for the public handpump are not visible because their values are extremely close to the point estimate. The most highly valued features associated with the largest increases in WTP over the reference scenario were access to private piped water connections (accessibility) and the addition of chlorination (water safety), followed by access to a public standpipe or handpump within 10 minutes of home (accessibility). Respondents also valued improvements to availability and reliability, but to a lesser extent.

https://doi.org/10.1371/journal.pwat.0000547.g004

The most highly valued features appeared to focus on accessibility and safety, consistent with our findings in the previous section. The largest increases in WTP over the reference scenario were associated with access to private piped water connections and the addition of chlorination (mean WTP increases: 29–39 GHS/m³, or 3.6-4.9 times higher than the reference scenario; p < 0.001). According to the model, on-plot piped water was valued a bit more highly than in-home connections (Fig 4). This unexpected result may be explained by respondents being accustomed to seeing on-plot connections in other communities or that activities such as food preparation and cooking commonly occur outside the main house in rural northern Ghana, which likely aligns well with on-plot services. The high preference for chlorination may reflect a considerable level of concern for the quality and safety of water. Notably, our mixed logit model showed that variations based on individual preferences were substantial for all three of these features (as shown by error bars in Fig 4, representing the spread of estimated individual WTP values). For example, while the mean WTP increase for chlorination was 36 GHS/m³ (4.5 times higher than the reference scenario), we estimated that 75% of respondents would be willing to pay only 18 GHS/m³ more than they would for unchlorinated water (2.3 times higher than the reference scenario).

Public standpipes (p = 0.007) and handpumps within 10 minutes of people’s homes (p = 0.013) were less valued than private piped connections but were still seen as improvements over reference handpumps that were 30 minutes away, with mean WTP increases of 10–12 GHS/m³ (1.3-1.5 times higher than the reference scenario; Fig 4). Notably, these two source types also exhibited much less variation in individual preferences. The random parameters associated with individual variations in these coefficient values were not statistically meaningful (p > 0.90), meaning that most respondents’ estimated WTP was very close to the mean. In contrast, respondents did not exhibit increased WTP for handpumps when collection times were 20 minutes rather than 30 minutes (p = 0.32).

Respondents also valued improvements to availability (mean WTP increase: 14 GHS/m³, or 1.8 times higher than the reference scenario; p < 0.001) and reliability (mean WTP increase: 7 GHS/m³, or 88% higher than the reference scenario; p = 0.011), although these features did not seem to be as critical as those related to accessibility and safety. WTP for availability and reliability features also varied considerably based on individual preferences (p = 0.007-0.009). For example, approximately 35% of respondents would not be willing to pay anything more to improve service reliability over the reference scenario, according to the WTP distribution estimated by our model (Fig 4). However, existing water service and respondent characteristics had minimal impact on individual variations in WTP for certain features, suggesting the importance of personal preference (Figure C in S1 Appendix).

In addition to these results from the baseline survey, we also conducted 424 supplemental DCE surveys in Northern Ghana focused specifically on small town residents with greater exposure to piped water services (see Table C in S1 Appendix for a comparison of these additional respondents with those from the baseline survey). The additional surveys applied the same approach used during the baseline, and they enabled us to investigate whether this slightly different context would alter how individuals valued the water service features included in the DCE scenarios.

Interestingly, results followed trends very similar to what we saw in the baseline data (Figure D in S1 Appendix). Private piped water and chlorination remained the most highly valued features, while public standpipes and handpumps within 10 minutes, as well as improved availability and reliability, were less valued but still seen as worthwhile enhancements over the reference scenario. We did observe some minor differences from these additional small town results. Private piped connections (especially on-plot) were valued more highly than chlorination, and public standpipes were valued more highly than handpumps. These differences may suggest that people’s appreciation of piped water becomes even more pronounced once they have seen or experienced it more directly. In most cases, WTP values for these improved services tended to be larger than those estimated from the original baseline surveys. Except for improved availability and public handpumps within 20 minutes, where mean WTP did not increase, mean WTP was 17–86% larger than in the original surveys. These differences may be associated with the fact that the supplemental small town respondents tended to be wealthier than our original respondents (Table C in S1 Appendix).

Finally, across both sets of DCE results, participants selected the “opt out” response 17% of the time (out of 3,860 total choice sets asked to 965 participants during baseline and supplemental surveys). Opting out means that the respondent did not want to choose either of the two scenarios presented in a choice set. Such a response may suggest that (i) the respondent is happy with their current water services or (ii) they were not willing to pay the prices associated with the available scenarios in the given choice set. Participants opted out most frequently (40% of the time) when both scenarios in the choice set were priced at the highest level (35 GHS/m³), which supports the second explanation. This result may suggest caution against setting tariff rates too high, even for the highest-quality services, as a substantial portion of our study population might be unwilling or unable to pay. We observed similar trends when considering each set of surveys individually (e.g., baseline respondents opted out in 16% of choice sets, while supplemental participants opted out 20% of the time).

4.1. Satisfaction levels and user preferences across studies and settings

Broadly, we found that approximately half of our respondents were satisfied with their existing water services, which may be somewhat higher than anticipated, since we targeted communities with relatively poor water and sanitation conditions. Based on responses concerning both existing services and hypothetical improvements, respondents seemed to value factors related to accessibility and safety most highly, although availability and reliability also played a role. A similar study conducted among communities in two districts of the Northern and Upper West regions in Ghana found that users were often dissatisfied with long collection times [6], suggesting similar accessibility concerns. In our case, while most respondents reported collection times below 30 minutes, DCE results suggest that users may still appreciate even shorter collection times (e.g., 10 minutes) if on-premises connections are not available.

With regard to water quality and safety, a study of urban water services in Chile found that perceived water quality had the largest impact on overall satisfaction [20], while residents in New Propontida, Greece exhibited high levels of WTP for improvements to drinking water quality and, to a lesser extent, for increased reliability [9]. Meanwhile, households in rural Ethiopia with access to improved water sources were more satisfied with the quality of water provided and, to a lesser extent, its availability, emphasizing that the source of water contributes to users’ perceptions of its quality [21]. These findings aligned with our result that respondents’ perceptions of water quality were highly associated with their satisfaction with existing services, with 68% of our satisfied respondents reporting quality/safety as one of the things they liked about their service (Fig 3). They also agreed with the high WTP for chlorination estimated from our hypothetical DCE scenarios. Despite demand for chlorination, careful implementation would be needed to translate these preferences into practice and avoid potential challenges related to chlorine dosing accuracy, consistency, and potential drop-off over time [22,23]. Community sensitization to sustain user-acceptability and overcome potential taste and odor concerns that could arise from chlorination would also be recommended [22,24].

Although our study and several of those discussed above suggest availability and reliability may be lower priorities for water users, other studies have noted their importance. For example, increased reliability and accessibility was associated with higher satisfaction in Ethiopia [7]. In a higher-income context, residents of Canberra, Australia were willing to increase their water bill to improve reliability through reductions in the frequency and duration of service interruptions [8]. Concerning availability, residents of urban informal settlements in Accra, Ghana exhibited higher WTP for water availability increasing from six to 12 hours per day, but increasing availability to 24 hours per day did not further increase WTP, suggesting that full availability may not be a high priority for some members of low-resource communities [25]. Another study found that users find intermittent water supplies less stressful when water service interruptions are scheduled and predictable. Predictable interruptions enable households to adapt by storing water in the home, which may allow them to have sufficient quantities available when needed (and meet the JMP availability criterion) [26]. In our study, while issues of reliability and availability may have been less critical for users, WTP results still suggested that they were valued. These previous studies confirmed that reliability should remain an important consideration when making service improvements.

Our results also suggested that female respondents were slightly less satisfied with their water services than men. Since women were primarily responsible for water collection in 93% of households, they may be more aware of water source availability and maintenance issues than men, and poorer water services may require more of women’s effort and time. A study focused on water-related gender disparities in rural central Ghana found that men and women had distinct experiences and perceptions regarding drinking water, which could influence differences in satisfaction. Women encountered daily accessibility challenges in their role as primary water source selectors and more commonly discussed concerns with water quality and safety [24]. However, it is notable that some of these concerns revolved around the taste and odor of chlorinated water, whereas we found high interest in chlorination. Ensuring that water treatment efforts do not create new issues is an important consideration, especially if negative perceptions around chlorination exist.

With regard to water service providers, Kumasi & Agbemor found that users were largely satisfied with services provided by WSMTs [6], whereas we found high levels of dissatisfaction when WSMTs provided water services. Among those users in the previous study who were dissatisfied with WSMTs, their leading causes for dissatisfaction involved a lack of communication and poor maintenance [6], similar to concerns noted by water users in another study in rural Ghana [24]. These items may provide insight into why our respondents served by WSMTs often tended to be dissatisfied. In an informal settlement within Accra, Ghana, residents reported a preference for water services managed by community-based committees or non-governmental organizations, likely due to low levels of public trust in the local water utility and metropolitan assembly, which manage existing water and sanitation services [27]. Meanwhile, households in rural Vietnam did not exhibit a clear preference for water services managed by the public or the private sector [28]. Taken together, these results suggest that specific context and local trust in existing institutions may matter more to user satisfaction than the general type of service provider (e.g., private, government, community-based) [27].

At a regional level, we found satisfaction levels in the Upper West and Upper East regions, and to a lesser extent in the North-East region, considerably lower than in the Northern region. Notably, these trends contrast with the 2021 Census, which reported that the Northern region was worse than the Upper West and Upper East in terms of basic service access and collection time. As noted previously, our study population was not a representative sample of these regions, as communities were selected specifically for poor existing water and sanitation conditions. Given that the Upper West and Upper East had higher water access levels in the census, the communities we sampled from those regions were likely worse off relative to their neighboring communities, potentially lowering satisfaction. In contrast, selected communities in the Northern region may have been more similar to their region as a whole. Our results provided some corroborating evidence for this explanation, as access to basic services in the Upper West (45%) was far below the Census value for rural residents in that region (71%). In the Northern region, our results (48%) were similar to the Census data (46%).

Finally, we found some potentially contradictory results regarding payments for water services. Those paying for water were less likely to be satisfied with their existing service, while satisfaction tended to be higher when a larger fraction of households in a community were making payments. Additionally, the affordability of existing services did not appear to be a major concern, and respondents seemed to be willing to pay more for better water service quality. This last finding runs somewhat contrary to a previous study in northern Ghana, which found resistance to paying for water because it is a basic need [6]. Although, that study also found that people were satisfied with service affordability, likely because most of them only paid when the system needed repairs, rather than paying directly for the water they collected. Our survey results suggested similar arrangements in many communities. This type of reactive maintenance payment scheme, as well as schemes based on fixed monthly or annual fees independent of water consumption, have been found to be less effective in keeping systems operational [6,29]. In contrast, payment schemes more focused on preventative rather than reactive maintenance would likely be important in maintaining acceptable levels of availability and reliability. We did not collect income information for all respondents, but Ghana’s most recent Living Standards Survey (GLSS 7, 2016–2017) estimated that mean annual household income in our study regions ranged from 13,000–23,000 GHS [4]. Ensuring that users see their payments as affordable and valuable would likely depend on equitable payments and whether users perceive that the money is being used effectively to maintain services.

4.2. Considering an additional rung on the JMP water service ladder: “Proximate access”

Based on our results from this context and the challenges of achieving full safely managed services in rural low-resource settings, we feel that it may be worthwhile to consider the addition of an intermediate access level that would fall between “basic” and “safely managed” on the current JMP water service ladder. Below, we evaluate possible definitions for this intermediate level, which we term “proximate access” as a preliminary suggestion.

A definition for this intermediate level could begin with the “basic” definition (an improved source with a collection time of 30 minutes or less), layering on certain “safely managed” criteria or making the “basic” definition more stringent to represent partial progress from basic to safely managed services. Using the data we have available concerning our baseline survey population, we were able to consider the impact of criteria related to accessibility (on-premises connections or shorter collection times) and availability (sufficient water is available when needed). Because water quality information was not available, we could not directly evaluate the quality criterion (“free from contamination”). However, we suggest that including this water quality criterion in the “proximate access” definition may be valuable, given that quality was a high priority for our respondents concerning both existing and hypothetical water services. For the remaining criteria, we evaluated how various possible definitions of “proximate access” would affect existing water service levels among our study population (Fig 5).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Water service ladder showing existing access levels among the baseline survey population (left), as well as satisfaction levels among those with basic and proximate access (right), using various possible definitions for the proposed proximate access level.

The satisfaction plot on the right focuses on basic and proximate access only because those with proximate access always represent a subset of respondents who were classified as having basic access under the existing service ladder. The vertical dashed line indicates satisfaction among those with basic access under the existing service ladder, to provide a point of comparison. The “safely managed” category represents the maximum possible percentage of the population at this level, because our dataset did not include information on water quality. Defining a new “proximate access” category using option d (access to an improved water source with a collection time within 10 minutes and for which households have sufficient water available when needed) shows the highest level of improvement in satisfied respondents compared to basic access.

https://doi.org/10.1371/journal.pwat.0000547.g005

Accessibility was highly valued by our respondents, but the small number of on-premises connections in our baseline study population confirmed that on-premises access can be challenging to achieve in remote rural settings. In conjunction, our DCE results suggested that respondents valued accessibility at three distinct levels. Improved sources within 10 minutes were valued more highly than those 20 or 30 minutes away, while on-premises connections were valued to a higher degree. This trend suggests to us a potential gap between accessibility criteria for basic access (which requires a collection time within 30 minutes) and safely managed access (which is strictly on-premises). Accordingly, we considered definitions of “proximate access” denoting access to an improved source that is (a) accessible on premises, or (b) within 10 minutes (but not on premises). Among our study population, the on-premises access criterion (a) was extremely restrictive, with <1% falling in the proximate access category (Fig 5). In contrast, the 10-minute collection time criterion (b) classified 21% of respondents as having proximate access, representing nearly half of those with basic access under the existing service ladder. In this study’s rural setting, this definition (b) may represent a substantial improvement toward safely managed services over basic services in terms of accessibility, while still not yet reaching the gold standard in full.

Respondents also valued availability and reliability, though to a lesser extent than accessibility and quality. Both availability and reliability play a role in the safely managed criterion that water is “available when needed,” so we also considered a third “proximate access” definition denoting (c) access to an improved source where water is available when needed. In our study context, this condition was frequently satisfied among our respondents, especially for those with at least basic access (79% of those with at least basic access reported sufficient water always being available, compared with 64% among those lacking basic access; p < 0.001). Accordingly, this definition (c) allowed most respondents with basic access in our study area under the existing service ladder to advance to the proximate category (Fig 5). At least in our context, this definition may be too lenient to produce a meaningful difference and improvement over basic access. Globally, approximately 1 billion people are served by intermittent water supply [30]. Based on the predictability and frequency of interruptions, some people with intermittent water supply may meet the “available when needed” safely managed criterion [26], though those not meeting it would also fail to meet the “proximate access” criterion. Users of intermittent water supply would likely need to experience improvements in predictability or continuity of water service to advance to “proximate access” under this definition.

Given these results, we considered combining the availability and 10-minute collection time criteria into a final possible definition (d). In this case, we classified 16% of our study population as having proximate access. Notably, this definition also resulted in the highest satisfaction level among the subset of respondents classified as having proximate access (Fig 5). Therefore, at least in the current context, we feel this definition (potentially with the added inclusion of the water quality criterion) would represent a meaningful service improvement over basic access that is feasible and valued by users. If adding the water quality criterion, we recommend monitoring efforts either use simplified test methods for fecal contamination developed and operationalized by Multiple Indicator Cluster Surveys (MICS) [31], or approximate water quality based on reported water treatment frequency or measured free chlorine residual levels.

To summarize, based on the characteristics of our study population in rural northern Ghana, we would recommend that a “proximate access” category could be defined as: access to an improved water source where collection time is 10 minutes or less and water is available when needed (and possibly free from contamination). Especially for rural low-resource contexts where on-premises access is a challenge, this proximate category could represent substantial progress most of the way toward safely managed services. Its inclusion in the overall service ladder could help increase motivation among local officials to continue making intermediary progress after basic access has been achieved, thereby making eventual transitions to fully safely managed services easier. However, in other, perhaps more urban settings, it is possible that the proximate category could discourage investments toward reaching safely managed services. Our results should be validated in other contexts before widespread implementation of a proximate category. Policy-makers and implementers would also need to be careful not to see proximate access as “good enough” and deprioritize further progress.

4.3. Study limitations

When considering the implications of our study findings, several limitations are important to note. First, our survey sample was dependent on baseline data collection for a local water and sanitation program, which focused on locations with the greatest water and sanitation needs. Accordingly, our study population does not provide a representative picture of these regions in Northern Ghana, and households in study communities may have had different expectations for water service improvements than those with higher service levels. Nevertheless, it does offer a window into conditions within critical settings where considerable progress toward safely managed services is needed. Additionally, while we attempted to randomly sample households using modified random walk procedures, it is possible this method introduced some sampling bias. However, daily review of satellite maps found good spatial coverage of household selection within communities and did not identify any bias in terms of households congregating around specific roads or points which could have introduced selection errors to this method [32], likely because these communities were open with many dirt paths between households to allow enumerators to move freely without needing to follow specific routes or roads. We also did not measure the quality of drinking water available to survey participants, which limited consideration of the “free from contamination” criterion for safely managed services. Relatedly, a study population characterized by a greater balance between different existing water source types and accessibility levels would have helped in characterizing trends related to these factors. However, our additional DCE surveys, conducted in small towns with greater exposure to on-premises piped connections, did suggest that our baseline results may be relevant across a broader set of contexts, particularly among somewhat wealthier populations with greater existing access to on-premises piped water in northern Ghana. Potentially, greater exposure to on-premises piped services may further increase the desire for this feature. Regardless, broader validation of our results in other settings would strengthen our recommendations around a “proximate access” category.

Additionally, multiple factors limited our quantitative consideration of district-level activities, such as the implementation of water governance accountability mechanisms and the promotion of participatory decision-making. First, we were unable to include some district indicators in our regression analysis due to collinearity with the region variable, so it is possible that the regional differences we observed may partially reflect distinct district-level practices. Second, certain activities, such as documenting service providers, making annual investment plans, and improving budget systems, were only practiced by one or two districts, which limited the ability of a regression analysis to capture any effects. Finally, our interviews with district officials suggested that resource challenges surrounded regular implementation of these activities, meaning that we may not have observed an effect even for activities we did include in the regression analysis. Therefore, while our regression model did not identify any district-level variables that were strongly associated with satisfaction levels, we do not believe this finding should discourage districts from pursuing such efforts. Even when local governments are not directly providing services, they still have important roles to play in terms of management, monitoring, and oversight.

Finally, our DCE analysis relied on hypothetical scenarios and did not ask participants to pay real money. Accordingly, our results reflect respondents’ stated WTP, which may overestimate revealed WTP (i.e., how much a respondent would be willing to pay when making a real purchase). In particular, a previous study focused on WTP for sanitation in northern Ghana found that respondents’ revealed WTP averaged 53–77% of stated WTP, depending on factors such as season and whether respondents were residents of small towns [19]. We did not consider seasonality in our current study, which could impact water availability and reliability as well as WTP, though we collected data during the dry season when water availability may be lower. Additionally, as most respondents in our current study were used to only paying sporadic contributions when repairs are needed, the DCE presented them with an unfamiliar scenario of paying volumetric tariffs, which could have led them to inaccurately estimate their WTP. Thus, while the trends ranking the relative value of different attributes revealed through the DCE model provide valuable insight, we feel that specific WTP values from the model should be treated with caution. For example, while they might provide motivation for setting or revising tariff rates for improved services, other information should be incorporated when deciding on the precise value of those rates. Other payment modalities not considered in this study, such as monthly flat fees, could also influence WTP and increase user preferences and water system revenue [33].