Water–sanitation–hygiene mapping: An improved approach for data collection at local level

Science of the Total Environment 463–464 (2013) 700–711

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Water–sanitation–hygiene mapping: An improved approach for data collection at local level Ricard Giné-Garriga a,⁎, Alejandro Jiménez-Fernández de Palencia a, Agustí Pérez-Foguet b a Research Group on Cooperation and Human Development, University Research Institute for Sustainability Science and Technology, Universitat Politècnica de Catalunya, Campus Nord, Edifici VX, Pl Eusebi Güell, 6, Barcelona, Spain b Dept. of Applied Mathematics III, Research Group on Cooperation and Human Development, University Research Institute for Sustainability Science and Technology, Civil Engineering School, Universitat Politècnica de Catalunya, Campus Nord, Edifici C2, c/Jordi Girona 1–3, Barcelona, Spain

H I G H L I G H T S • • • • •

We present an integrated method for WASH-related data collection at local level. The survey design combines a waterpoint mapping and a household survey. Simple statistical analysis validates the data from the viewpoint of decision-making. Data provide policymakers with evidences to inform planning and targeting processes. We conclude that integrated data collection mechanisms can be designed to support local policymaking.

a r t i c l e

i n f o

Article history: Received 2 December 2012 Received in revised form 22 May 2013 Accepted 2 June 2013 Available online 10 July 2013 Editor: Simon James Pollard Keywords: Data collection Data management Water point mapping Household survey WASH East Africa

a b s t r a c t Strategic planning and appropriate development and management of water and sanitation services are strongly supported by accurate and accessible data. If adequately exploited, these data might assist water managers with performance monitoring, benchmarking comparisons, policy progress evaluation, resources allocation, and decision making. A variety of tools and techniques are in place to collect such information. However, some methodological weaknesses arise when developing an instrument for routine data collection, particularly at local level: i) comparability problems due to heterogeneity of indicators, ii) poor reliability of collected data, iii) inadequate combination of different information sources, and iv) statistical validity of produced estimates when disaggregated into small geographic subareas. This study proposes an improved approach for water, sanitation and hygiene (WASH) data collection at decentralised level in low income settings, as an attempt to overcome previous shortcomings. The ultimate aim is to provide local policymakers with strong evidences to inform their planning decisions. The survey design takes the Water Point Mapping (WPM) as a starting point to record all available water sources at a particular location. This information is then linked to data produced by a household survey. Different survey instruments are implemented to collect reliable data by employing a variety of techniques, such as structured questionnaires, direct observation and water quality testing. The collected data is finally validated through simple statistical analysis, which in turn produces valuable outputs that might feed into the decision-making process. In order to demonstrate the applicability of the method, outcomes produced from three different case studies (Homa Bay District –Kenya–; Kibondo District –Tanzania–; and Municipality of Manhiça –Mozambique–) are presented. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Abbreviations: JMP, Joint Monitoring Programme for Water Supply and Sanitation; MICS, Multiple Indicator Cluster Survey; NGO, Non-governmental organization; UNICEF, United Nations Children's Fund; WASH, Water, Sanitation and Hygiene; WP, Waterpoint; WPM, Water Point Mapping. ⁎ Corresponding author at: Research Group on Cooperation and Human Development, University Research Institute for Sustainability Science and Technology, Universitat Politècnica de Catalunya, Spain. Tel.: +34 405 43 75. E-mail addresses: [email protected] (R. Giné-Garriga), [email protected] (A.J.-F. de Palencia), [email protected] (A. Pérez-Foguet). 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.06.005

Water and sanitation improvements together with good hygiene (WASH) produce evident effects on health population (Cairncross et al., 2010; Curtis and Cairncross, 2003; Esrey et al., 1991; Feachem, 1984; Fewtrell et al., 2005). However, universal access to safe drinking water and basic sanitation remains a huge challenge in many low income countries (Joint Monitoring Programme, 2012a), where vast numbers of people are not properly provided for by these basic services. To help end this appalling state of affairs, the sector has

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

been facing a gradual process of decentralisation, where the responsibility in service provision moves to local authorities. It is believed that decentralised governments have an informational advantage over the central government with regard to local needs and priorities, for which reason they are assumed to supply services in accordance with demand, allocate resources more equitably, and ultimately conceive and implement policies with a focus on poverty reduction (Crook, 2003; Devas and Grant, 2003; Steiner, 2007). To effectively do this, local governments need to make evidence-based decisions, which primarily depend on the availability of accessible, accurate and reliable data that are routinely collected, disseminated and updated. Amongst others, these data may be employed to i) measure progress and performance, ii) improve transparency in budgetary procedures and promote increased investments in the sector, and iii) allocate resources to deliver services where they are most needed. Today, reliable information on key indicators at local level often lacks, but even when it is available, the uptake for such data by policymakers is, at best, challenging (WaterAid, 2010). Limited capacities of recipient governmental bodies, inadequate sector-related institutional framework, and lack of data updating mechanisms are common reasons that hamper an adequate appropriation and continued use of the data for planning and monitoring purposes (Joint Monitoring Programme, 2011; World Health Organization, 2012). In an effort to address one of the shortcomings cited above, i.e. the lack of reliable data, this study deals with the design of adequate methodologies for routine data collection. A variety of tools and techniques have been developed in recent years to collect primary data for the WASH sector. Amongst others, the Water Point Mapping –WPM– (WaterAid and ODI, 2005), the UNICEF-supported Multiple Indicator Cluster Survey –MICS– (United Nations Children's Fund, 2006), the Rapid Assessment of Drinking Water Quality –RADWQ– (Howard et al., 2003, draft), and the Water Safety Plans (Bartram et al., 2009). However, methodological problems arise when they are implemented at local scale to produce reliable inputs for planning support. First critical shortcoming is related to the type of data required to monitor the sector, since different information sources may be required (Joint Monitoring Programme, 2012b). Household surveys are by large the most commonly used tools for collecting WASH data (Joint Monitoring Programme, 2006; Macro International Inc., 1996; United Nations Children's Fund, 2006). But a focus on households is not sufficient to answer many relevant questions, and hence needs to be supplemented with data from other sources. For instance, an audit at the water point might provide insight into operational and management-related aspects of the service. A methodology to efficiently combine these two types of information sources should have potential for wider implementation. Another key limitation is that of comparability (Joint Monitoring Programme, 2006), since a variety of indicators are being simultaneously employed to measure different aspects of the level of service. More often than not, to assess trends over periods of time or to compare indicators regionally has therefore remained challenging. As a first step against this comparability problem, the Joint Monitoring Programme for Water Supply and Sanitation (JMP) formulated a set of harmonized survey questions (Joint Monitoring Programme, 2006) to provide worldwide reliable estimates of drinking-water and sanitation coverage at national level (Joint Monitoring Programme, 2012a). In so doing, JMP has improved the processes and approaches to monitoring the sector, though the definitions employed have been criticised as being too infrastructure-based. (Giné-Garriga and Pérez-Foguet, under review-b; Giné-Garriga et al., 2011; Hunt, 2001; Jiménez and Pérez-Foguet, 2012). Today, an ongoing consultative process is debating a consolidated proposal of targets and indicators for the post-2015 monitoring framework (Joint Monitoring Programme, 2011; Joint Monitoring Programme, 2012c). The techniques employed for data acquisition also play a key role in terms of data reliability and validity (United Nations Children's Fund, 2006). A well-designed questionnaire helps elicit a response that is

701

accurate and measures the things one seeks to measure. On the other hand, interviews with predetermined and closed-end questions are not conducive to study respondent's perceptions or motivations (Grosh, 1997), thus pointing out the need for employing alternative survey instruments to avoid bias in survey's outcomes. For instance, water quality should be bacteriologically tested (Howard et al., 2003, draft; Jiménez and Pérez-Foguet, 2012; Joint Monitoring Programme, 2011); while study of handwashing through structured observation may help avoid over-reporting of “desirable” hygiene behaviours (Manun'Ebo et al., 1997). Finally, there is an issue with the statistical precision of the estimates. A common monitoring need in local decision-making is to assess separately the performance of the lowest administrative subunits (e.g. communities, villages, etc.) in the area of interest (e.g. district, municipality, etc.). Since the number of these administrative subunits is generally large, the level in which information needs to be disaggregated is high, and one is therefore faced with the need to balance precision against cost when deciding the size of the sample (Bennett et al., 1991; Grosh, 1997; Lwanga and Lemeshow, 1991). Moreover, a scientifically valid sampling methodology is necessary to achieve reliable estimates. For household surveys, a cluster sampling design has proved a practical solution (Bennett et al., 1991; Lemeshow and Stroh, 1988; United Nations Children's Fund, 2006). And water point mapping exercises, where a comprehensive record of water sources is undertaken (i.e. no sampling), have also been successfully implemented to monitor the distribution and status of water supplies (WaterAid, 2010). In sum, a need for further research into feasible alternatives for data collection to the currently used strategies has been highlighted (Joint Monitoring Programme, 2011), and the purpose of this study is to present a new specific approach for the WASH sector at local level, as an attempt to overcome previous shortcomings. It takes the WPM as a starting point to record all available water sources at a particular location, which results in the need of covering the whole area of intervention. This information is then combined with data provided from a household-based survey, in which a representative sample of households is selected to assess sanitation and hygiene habits. In brief, taking advantage of the current momentum of WPM as field data collection method in the water sector (Government of Liberia, 2011; Government of Sierra Leone, 2012; Jiménez and Perez-Foguet, 2011; Pearce and Howman, 2012; WaterAid, 2010) and the growing interest among development stakeholders in harmonizing sector monitoring (Joint Monitoring Programme, 2012c), this study suggests a cost-efficient alternative to simultaneously perform a WPM together with a household survey, thus producing a comprehensive WASH database as a valuable output for policymaking. To test the applicability and validity of the proposed approach, three different case studies in East Africa are presented. In Section 2, basic concepts of the evaluation framework employed in this study are outlined. The methodology proposed to collect WASH primary data is described in Section 3. It presents the three case studies and highlights key features of the approaches adopted in each one of them. Section 4 computes statistical validity of the method and, in so doing, provides useful guidelines on data exploitation for decision-makers. Integral to this discussion there are a variety of alternatives to disseminate achieved results, in an effort to provide clear and accurate policy messages. The paper concludes that efficient data collection mechanisms can be designed to produce reliable estimates for local planning processes. Their implementation in the real world, however, is to a certain extent elusive; and specific challenges that remain unaddressed are pointed out as ways forward. 2. Evaluation framework This section introduces core aspects of the evaluation framework proposed to locally assess the WASH status. First, the two methodologies for

702

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

data collection in which we base our approach are presented, i.e. the Water Point Mapping (WPM) and the Multiple Indicator Cluster Survey (MICS). Second, it discusses the issue of the sample size, as the survey design has to enable the compilation of accurate primary data to produce statistically representative estimates. Third, a reduced set of measurable indicators is proposed as the basis of the monitoring strategy. 2.1. The Water Point Mapping Mapping of water points has been in use by NGOs and agencies worldwide for over a decade, particularly in sub-Saharan Africa (e.g. Malawi, Tanzania, Ghana, Ethiopia, Zambia, Liberia, Sierra Leone, etc.). This methodology, largely promoted by the NGO WaterAid, can be defined as an ‘exercise whereby the geographical positions of all improved water points1 in an area are gathered in addition to management, technical and demographical information’ (WaterAid and ODI, 2005). WPM involves the presentation of these data in a spatial context, which enables a rapid visualization of the distribution and status of water supplies. A major advantage is that water point maps provide a clear message on who is and is not served; and particularly in rural areas, they are being used to highlight equity issues and schemes' functionality levels at and below the district level. This information can be employed to inform decentralized governments about the planning of investments to increase water coverage (Jiménez and Pérez-Foguet, 2010; WaterAid, 2010). Specifically, the mapping does not refer to a fixed set of indicators, and two different actions are suggested in this regard: i) biological testing to ensure water quality; and ii) the inclusion of unimproved sources. First, water quality analysis has long been nearly absent from water coverage assessments because of affordability issues (Howard et al., 2003, draft; Joint Monitoring Programme, 2010). In the absence of such information, it is assumed that certain types of water supplies categorized as ‘improved’ are likely to provide water of better quality than traditional unimproved sources (Joint Monitoring Programme, 2000; Joint Monitoring Programme, 2012a). This assumption, though, appears over-optimistic, and improved technologies do not always deliver safe water (Giné Garriga and Pérez-Foguet, under review-b; Jiménez and Pérez-Foguet, 2012; Sutton, 2008). Contrary to what might be expected, and particularly in comparison with overall investments projected for new infrastructure or with ad hoc quality testing campaigns, water quality surveillance does not significantly impact on the overall cost of the mapping exercise: from USD 12 to 15 dollars/waterpoint in standard WPM (Stoupy and Sugden, 2003) up to USD 20 when quality testing is included (Jiménez and Pérez-Foguet, 2012). Second, being the original focus of WPM on improved waterpoints, unimproved sources may be also mapped if they are accessed for domestic purposes. A thorough analysis of collected data would shed light on the suitability of the improved/unimproved classification proposed by the JMP, but more importantly, this would help understand equity issues in service delivery (Giné-Garriga and Pérez-Foguet, under review-b; Jiménez and Pérez-Foguet, 2011; Joint Monitoring Programme, 2012a).

(i.e. all enumeration areas as communities, villages, etc.). Taking advantage of this logistic arrangement, and in addition to the mapping, a household-based survey may be thus designed to evaluate sanitation and hygienic practices at the dwelling. As it may be assumed that all households are located within walking distance of one water source (either improved or unimproved), the approach adopted practically ensures full inclusion of families in the sampling frame. In terms of technique, the design and selection of the sample draws on the MICS, i.e. a methodology developed by UNICEF (United Nations Children's Fund, 2006) to collect social data, which is ultimately required amongst others for monitoring the goals and targets of the Millennium Declaration or producing core United Nations' development indices. The study population is stratified into a number of small mutually exclusive and exhaustive groups, so that members of one group cannot be simultaneously included in another group. In this study, however, main difference is that when sampling, a sample of households is selected from each stratum (stratified sampling), rather than selecting a reduced number of strata, from which a subsample of households is identified (cluster sampling). In so doing, the risk of homogeneity within the strata remains relatively low, thus reducing the need for applying any correction factor in sample size determination, i.e. the “design effect2”. A “design effect” of 1 is accepted in stratified random sampling, though ten-fold or even higher variations are not uncommon values in cluster samplings with large cluster's sizes (Kish, 1980). In a WASH cluster survey, a value of 4 may be appropriate as acknowledged by the United Nations Children's Fund (2009). 2.3. Sample size and precision In local decision-making, of interest is the evaluation of the level of service for the recipient administrative unit as a whole. However, acknowledging that administrative subunits may have uneven coverage, there is also concern for estimating their performance to identify the most vulnerable areas. In other words, one regional coverage value might be sufficient from the viewpoint of central governments; but since such value says nothing about local variations, estimates at the lowest administrative scale are required for decentralised planning. To produce local robust estimates substantially increases the required size of the sample, which directly affects the cost of the survey. The goal of WPM is to develop a comprehensive record of all water points available in the area of intervention. There is thus no need of sampling. For the household survey, in contrast, a statistically representative sample needs to be selected. The basic sampling unit is the household, and the size of a representative sample n is numerically given by Cochran (1977, third edition): n¼

α

p

D 1

The types of water points considered as improved are consistent with those accepted internationally by the WHO/UNICEF Joint Monitoring Programme Joint Monitoring Programme. Core questions on drinking-water and sanitation for household surveys. WHO/UNICEF, Geneva/New York, 2006. More specifically, an improved water point is a place with some improved facilities where water is drawn for various uses such as drinking, washing and cooking Stoupy O, Sugden S. Halving the Number of People without Access to Safe Water by 2015 – A Malawian Perspective. Part 2: New indicators for the millennium development goal. WaterAid, London, 2003.

d2

ð1Þ

where:

2.2. Household survey A major strength of WPM is, per definition, comprehensiveness with respect to the sample of water points audited, which entails complete geographic representation of all strata in the study area

z21−α=2 pð1−pÞD

is the confidence level, and z is a constant which relates to the normally distributed estimator of the specified level. For a confidence level of 95% (α = 0.05), the value of z1 − α/2 is 1.96 (z1 − α/2 = 1.64 when α is 0.1; z1 − α/2 = 1.28 when α is 0.2); is the assumed proportion of households giving a particular response for one given question. The “safest” choice is a figure of 0.5, since the sample size required is largest when p = 0.5; is the sample design effect. As mentioned, D = 1 in stratified random sampling. However, acknowledging that a complete random exercise for household selection is almost

2 The “design effect” is an adjustment that measures the efficiency of the sample design, and is calculated by the ratio of the variance of an estimator to the variance of the same estimator computed under the assumption of simple random sampling.

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

unachievable in each stratum, a value of 2 is recommended. It is noteworthy that in comparison with the sampling plan required in a standard cluster survey, where D = 4 (United Nations Children's Fund, 2009), the sampling approach adopted in this study halves the sample size, which considerably reduces the overall cost of the data collection exercise; and is the required precision on either side of the proportion. A typically used figure in similar surveys is d = ± 0.05, based on the argument that lower precision would produce unreliable results while a higher precision would be too expensive as it would require a very large survey. This precision may be considered at highest scale of intervention. Estimates at lower administrative scale should be assessed with lower precision; i.e. d = ± 0.10 or ±0.15.

d

As an example, a minimum sample size n of 192 would be required to produce estimates in each administrative subunit within 20% (±10%) of the true proportion with 95% confidence (D = 2). If the sample design effect is estimated as 4, twice the number of individuals would have to be studied (i.e. 384) to obtain the same precision. From previous figures, however, it can be seen that Eq. (1) is valid where populations are at least medium-size (N > 100). In contrast, when applied to more reduced populations, it produces unachievable figures or large sampling errors. For use in local household-based surveys, Giné-Garriga and Pérez-Foguet (under review-a) proposed an alternative approach to determine the sample size, in which the practitioner may easily identify the sampling plan that best balances precision and cost (Table 1).

2.4. Water, sanitation and hygiene indicators A core element of any evaluation framework is the set of indicators in which base the analysis. It is evident that from a WASH perspective, a wide range of variables exists to assess the current status of service level. Of particular interest is the recently adopted human rights framework, which reflects the concept of progressive realization in the level of service and requires the definition of specific indicators to deal with the issues of affordability, quality, reliability and non-discrimination, amongst others; or the debate guided by WHO and UNICEF about the post-2015 monitoring of WASH (Joint Monitoring Programme, 2011, 2012c). To exactly identify what should be measured remains challenging, though, and rules of thumb for the selection process include the following criteria. First, in terms of efficiency, the number of indicators should be as reduced as possible but sufficient to ensure a thorough description of the context in which the service is delivered (Joint Monitoring Programme, 2011; United Nations Children's Fund, 2006). Second, to resolve the comparability problems, survey questions need to be harmonized with those internationally accepted (Joint Monitoring Programme,

703

2006; Joint Monitoring Programme, 2012c). Finally, indicators should be EASSY (Jiménez et al., 2009): Easily measurable at local level, Accurately defined, Standardized and compatible with data collected elsewhere, Scalable at different administrative levels, and Yearly updatable. In Table 2, a short list of core indicators is summarized. Moreover and beyond average attainments, it is accepted that any evaluation framework should identify the high-risk groups in which policy-makers may prioritize efforts and resources (Joint Monitoring Programme, 2012c). The concern is to identify gaps in WASH outcomes between the poor and the better off. To do this, one option is to employ “direct” measures of living standards, such as household income or expenditure, though they are often unreliable (Filmer and Pritchett, 2001). Another approach is to use a “proxy” measure, such as a wealth index constructed from information on household ownership of durable goods (Booysen et al., 2008; Filmer and Pritchett, 2001; O'Donnell et al., 2007), education level of household-head, sex of household head, etc. These data promote evidence-based pro-poor planning and targeting processes, and ultimately help improve service level of the most vulnerable.

3. Method As abovementioned, the approach adopted for data collection combines a mapping of water sources with a stratified survey of households. Different methodologies exist which combine the waterpoint and the household as key information sources, but they commonly differ from the method proposed herein in i) the focus – national rather than local–, and in ii) the statistical precision of the estimates –inadequate to support local level decision-making–. Key features of the proposed methodology include i) an exhaustive identification of enumeration areas (administrative subunits as communities, villages, etc.); ii) audit in each enumeration area of all improved and unimproved water points accessed for domestic purposes; and iii) random selection of a sample size of households that is representative at the local administrative level (e.g. district, municipality, etc.) and below. The mapping of waterpoints is exhaustive regardless functionality issues, though the inclusion of unimproved sources in the analysis will be dependent on the scope of the exercise and available resources. The need to tackle equity issues has been highlighted from the viewpoint of human rights, and any exercise covering unimproved waterpoints would provide inputs to elucidate the access pattern of the population. In rural contexts, however, this type of water source may be common, thus increasing significantly the budget and resources devoted to data collection in case of inclusion. Similarly, where main water technology is piped systems with household connections, the idea of a comprehensive audit of all these private points-of-use is practically impossible. A more convenient solution would be to visit the distribution tank and

Table 1 Sample size n for different values of N, α and d. Source: Giné-Garriga & Pérez-Foguet (under review-a). N

8 10 15 20 25 50 75 100 150 250 500 Eq. (1)

α = 95%

α = 90%

α = 80%

d b 0.1

d b 0.15

d b 0.20

d b 0.25

d b 0.1

d b 0.15

d b 0.20

d b 0.25

d b 0.1

d b 0.15

d b 0.20

d b 0.25

– – 14 18 22 36 46 53 64 75 87 96

– 9 13 16 18 26 30 33 37 40 43 43

7 8 11 13 14 18 21 22 23 24

6 7 9 10 11 13 14 15

– 9 12 14 16 22 25 27 29

7 8 10 11 12 15 17 17

6 7 8 9 9 11

– – 14 17 19 28 32 35 39

7 9 11 13 14 18

6 7 9 10 10

5 6 7 7

24

15

– – 14 18 21 33 40 46 53 60 67 67

30

17

11

41

18

10

7

704

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

Table 2 List of core WASH indicators. WASH component

Indicator

Rationale

Source of information

Technique

Water supply

Access to improved water sourcesa, b

% households with access to improved water supply

Household

Direct questioning

Waterpoint

Visit to all waterpoints

One way distance to water source (km) a, b

% of households adequately covered (based on the standard source:man ratio) % of households spending, on average, more than 30 min in fetching water

Household

Direct questioning

Individual collecting watera

% households in which women shoulder the burden in collecting water

Household

Direct questioning

Domestic water consumption

Average rate of per capita domestic water consumption (based on of the number of containers consumed per day and the rough volume of these containers) % functional water points

Core water-related indicator. An improved source serves as a proxy indicator for whether a household's drinking-water is safe. To geographically show the least covered administrative subunits, i.e. with less number of water points compared to the population living there. To assess whether the source is sufficiently close to the household to ensure an adequate daily volume of water for basic domestic purposes. It also help determine the saving in time of fetching water, as a major expected benefit from the user's side. This information helps identify gender and generational disparities with respect to water-hauling responsibilities. It also ascertains who would profit from bringing water closer to households. Distance to the water source may be an indirect indicator of water use, but it is not accurate enough to draw conclusions. From the health viewpoint, it is important to determine whether the volume of water collected for basic needs reaches the minimum target value. To highlight sustainability issues, i.e. to identify operation and Maintenance (O&M) problems and to assess the overall quality of the O&M system. To evaluate water safety, specifically to determine presence of faecal coliforms and few other critical parameters (pH, conductivity, turbidity and nitrates) To identify seasonal or intermittent supplies, and to help assess reliability of the service. A water point is considered to be seasonal if a seasonal interruption in the supply of more than one month is reported. Where seasonality is high, people often need to search for alternative sources during dry season A key sustainability aspect of the supply. For successful and sustainable water schemes management, a proper institutional setting is required, and at least functional water user committees need to be established. A key sustainability aspect of the supply. Access to skills and spares promotes locally-based maintenance. A key sustainability aspect of the supply, particularly related to improve transparency and accountability. Regular meetings are proxies of upward and downward accountability, both on part of the local authority when dealing with the water committee, and on the latter when tackling public issues with beneficiaries A key community crosscutting issue. Fundamental human rights criteria include accessibility, affordability and non-discrimination. Core water-related indicator. Important to check the current use of the facility, rather than mere household's ownership of the toilet. To help distinguish between open defecation and latrine sharing, since both practices are categorized as unimproved in JMP figures. The shared status of a sanitation facility may entail poorer hygienic conditions than facilities used by a single household. Regardless the category of the toilet, it is important to determine whether the maintenance of a facility undercut its hygienic quality and jeopardize a continued use. Four proxies are verified: i) inside cleanliness, ii) presence of insects, iii), smell and iv) privacy. A rigorous assessment of handwashing behaviour would entail structured observation –a prohibitively expensive exercise–. An assessment of the adequacy of handwashing facilities, i.e. presence of soap and water, may be an in-between solution.

Household

Direct questioning/ observation

Waterpoint

Observation

Waterpoint

Waterpoint

Water quality testing (portable kit) Direct questioning

Waterpoint

Direct questioning

Waterpoint

Direct questioning

Waterpoint

Direct questioning

Waterpoint/ household

Direct questioning

Household

Observation

Household

Observation

Household Household

Direct questioning/ observation Observation

Household

Observation

Household

Direct questioning/ observation Direct questioning

Operational status of water sourceb Water quality (bacteriological contamination)b Seasonality of water resourcesb

% bacteriological acceptable water sources

Management systemb

% facilities with a functional and registered water committee

Maintenance system

% facilities with local access to technical skills and spare parts % of facilities in which at least 1 meeting was held during last year to discuss income and expenditure (both with the community and the local authority)

Financial control

Sanitation

% year-round water sources

Pro-poor service delivery

% water entities which exempt vulnerable houses from paying for water

Access to and use of sanitation

% households with access to improved sanitationa, b % households practicing open defecationb

Latrine conditions

Hygiene

a b

% households sharing improved sanitation facilitiesa % latrines maintained in adequate sanitary conditions

Handwashing deviceb

% latrines with appropriate handwashing device

Point-of-use water treatmenta Disposal of children's stoolsa, b

% households with adequate water treatment % child caregivers correctly handling baby excreta

Children's faeces are the most likely cause of faecal contamination to the immediate household environment.

JMP indicator. Indicator proposed by JMP for Post-2015 monitoring of drinking-water, sanitation and hygiene.

Household

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

a reduced number of domestic taps, which are taken as representative of the overall system. The household survey is conducted in parallel with the mapping. Ideally, a defined number of households will be selected in a statistically random manner from a comprehensive list of all households in the subunit of study. However, such a list does often lack. Then, if the population size is small, the optimum alternative may be to create a list by carrying out a quick census. In those cases where enumerating all households is impracticable, literature suggests different sampling techniques to achieve a random or near-random selection (Bennett et al., 1991; Frerichs and Tar, 1989; Lemeshow and Stroh, 1988). They usually involve two stages: the identification of one or various households to be the starting point, and a method for selecting “n” successive households, preferably spread widely over the community. In the end, where a complete random exercise is not achievable, any methodology during the sampling process which promotes that the sample is as representative as possible would be acceptable, as long as it is clear and unambiguous, and does not give the enumerator the opportunity to make personal choices which may introduce bias. In these cases, however, and to ensure data validity, to apply a correction factor in sample size determination (D = 2) is recommended. In terms of technique, the method relies on a variety of mechanisms to assure quality of produced outcomes. Among the most important are: – Territorial delimitation of study area. As an exercise to support planning, administrative subunits in which base data collection should play a relevant administrative role in decentralised service delivery. Thus, they should be adequately delimited, unambiguous and well-known by both decision-makers and local population. – Design of survey instruments. On the basis of a reduced set of reliable and objective indicators (Table 2), appropriate survey tools should be developed for an accurate assessment of the WASH status. This study is reliant on a combination of quantitative and qualitative study tools, which are specially designed to collect data from the water point and the household. Field inspections at the source employ a standardized checklist to evaluate the existence, quality and functionality of the facility; and a water sample is also collected for on-site bacteriological testing. At the dwelling, information related to service level is captured through a structured interview administered to primary care-givers. In addition, direct observation enables a complementary evaluation of domestic hygiene habits that may not be otherwise assessed, as sanitary conditions of the latrine, existence and adequacy of the handwashing facility, etc.. – Involvement and participation of local authorities. This study engages in various stages of the process with those government bodies with competences in WASH. Specifically in data collection, the commitment of officers belonging to the local government i) helps ensure a link between field workers and the local structures at community level, and ii) promotes sense of ownership over the process, as prerequisite for incorporating the data into decision-making. As important of promoting collaborative data collection methods is to foresee the viability of future data update activities, and accessibility and reliability of information have been two core criteria when preparing the survey instruments. Moreover, a consultative approach has been adopted for indicators' definition to tailor the survey to each particular context. Finally, data collection focuses on the administrative scale in which decisions are based, thus producing relevant information for local policy-makers. – Pilot study. A pilot run helps explore the suitability of the approach adopted, i.e. methodology and study instruments. Further finetuning (question wording and ordering, filtered questions, deletion of pointless questions, etc.) follows the pilot. – Data processing: The data entry process needs to be supervised, and the produced datasets need to be validated on a regular basis. Various

705

quality control procedures must be in place to ensure that the data reflects the true position as accurately as possible, and routine analysis of database or random checks of a reduced number of questionnaires may help detect data inconsistencies and improve database robustness. 3.1. Study area Three different East African settings were selected as initial case studies to test the applicability and validity of the proposed methodology, namely the district of Kibondo (Tanzania, in 2010), the district of Homa Bay (Kenya, in 2011) and the municipality of Manhiça (Mozambique, in 2012). The implementation of each case study adopted particular features, which are briefly summarized in Table 3, and scope of work was designed on the basis of local needs (e.g. inclusion/exclusion of unimproved waterpoints, visit to schools and health centres, the focus and level of detail required in survey questionnaires, etc.). However, they all shared same approach, method and goals: i) they were formulated against specific call from a development-related institution to support local level decision-making (in Tanzania, a Spanish NGO; in Kenya, UNICEF; and in Mozambique, the Spanish Agency for International Development Cooperation); ii) the Research Group on Cooperation and Human Development at the Technical University of Catalunya undertook overall coordination of the study; iii) the local authority was engaged as principal stakeholder throughout the process; and iv) a consultancy firm was contracted for field work support. 4. Discussion In previous section, a simplified approach to survey design for WASH primary data collection has been outlined. The goal of the discussion first focuses on providing statistical robustness of the methodology. To do this, we compute basic statistical parameters, in which also base the definition of criteria that will help validate the collected data from the viewpoint of decision-making. Second, and with the aim of communicating clear messages to policy-makers, a number of alternatives to disseminate achieved results are presented. 4.1. Estimating the precision of a proportion The data collected at the dwelling, because of the sampling strategy employed for households' selection, require statistical validation. The ultimate goal is to guarantee reliability of any outcome produced and thus avoid decisions based on false or misleading assumptions. To this end, estimates of proportions may be calculated together with precision of those estimates, so that confidence intervals can be assessed. As shown, all calculations described below are simple and can be easily computed in any standard spread-sheet. The proportion p, for example, of households in the ith subunit with access to improved sanitation is given by Eq. (2):

pi ¼

yi ni

ð2Þ

where: yi ni

number of households in the ith subunit with access to improved sanitation; and number of surveyed households in the ith subunit.

When estimating the proportion at the overall administrative unit, population size of subunits should be taken into consideration to avoid subunits' under or overrepresentation. Therefore, achieved responses should be weighted in proportion to the actual population of

706

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

Table 3 Key features of the approach adopted for data collection in each case study. Case study

Cost, in USDa, b

Adm. division Unit (subunits)

Kibondo, Tanzania

Data collection

Key features

No. WPsc

No. HH 3.656

District (20 wards)

13.578 P (42%), T (45%) and OC (13%)

986 IWPs

Homa Bay, Kenya

District (5 divisions)

32.389 P (74%), T (17%) and OC (9%)

255 187 IWPs and 68 UWPs

1.157

Manhiça, Mozambique

Municipality (18 bairros)

23.719 P (41%), T (42%) and OC (17%)

228 224 IWPs and 4 UWPs

1.229

− The total area is 16,058 km2 and the population is estimated at 414,764 (2002 Tanzania National Census). − Sampling Plan (at ward level): α = 0.05; D = 2; d = ±0.10; n (min) = 192. − Unimproved WPs were not audited. The WP audit included 38 questions (30 min per WP) + 1 water quality test. − HH checklist included 18 questions related to sanitation and domestic hygiene issues (10 min per HH). − The field team included one staff from Spanish NGO, 1 technician from District Water Department, two staff from a consultancy firm and two people from each visited village. Field work was completed in 42 days. − The total area is 1,169.9 km2, and the total population is about 366,620 (2009 National Census). − Sampling Plan (at division level): α = 0.05; D = 2; d = ±0.10; n (min) = 192. − Unimproved WPs were audited in only 3 out of 5 divisions. The WP audit included 38 questions (30 min per WP) + 1 water quality test. − HH checklist included 65 questions related to water, sanitation and domestic hygiene issues (35 min per HH). − Data collection did not include urban areas. It included schools (85) and health centres (37). − The field team included tree staff from GRECDH–UPC (1 fully involved), 1 technician from the District Water Department (partially involved), 1 technician from the District Public Health Department (partially involved), 8 staff from a consultancy firm, and one people from each visited community. Field work was completed in 33 days. − The total area is 250 km2 and the population is estimated at 57,512 (2007 national estimates) − Sampling Plan (at bairro level): α = 0.05; D = 2; d = ±0.15; n (min) = 86. Field work was completed in 39 days. − Audit of improved and unimproved WPs. The WP audit included 41 questions (30 min per WP) + 1 water quality test − HH checklist included 82 questions related to water, sanitation and domestic hygiene issues (45 min per HH) − Data collection included schools (16) and health centres (2) − The field team included three staff from GRECDH–UPC (1 fully involved), 3 technicians from the Vereação para Urbanização, Construção, Água e Saneamento (partially involved), 14 staff from a consultancy firm and 1 people from each visited village. Field work was completed in 29 days.

a Includes data collection and data entry into the database. It does not include the cost of the portable kit for water quality analysis and consumables. In percentage, overall budget broken down into personnel, transport, and others. b Type of costs includes P for personnel; T for Transport; and OC for Other costs. c Type of waterpoints includes IWP for Improved waterpoint and UWP for unimproved waterpoint.

each subunit. To compute the confidence limits for pi, we use the F distribution (Leemis and Trivedi, 1996), i.e. the so called Clooper–Pearson interval (Reiczigel, 2003):

pL ¼

1þ

!−1

n−y þ 1

ð3:1Þ

y F 2y;2ðn−yþ1Þ;1−α 2

pU ¼

1þ

n−y ðy þ 1Þ F 2ðyþ1Þ;2ðn−yÞ;α

!−1 ð3:2Þ

2

where: pu and pl are the upper and lower limits of the confidence interval; and α is the confidence level. In this exercise, two different confidence levels have been employed for calculation purposes, i.e. 90% (α = 0.1) and 80% (α = 0.2). Summary of aforementioned statistics (proportion and confidence interval) for the survey variables (listed in Table 2) are presented in Tables 4 and 5 –Kenya–, 6 –Tanzania– and 7 –Mozambique–, though

on practical grounds, estimates of only few indicators are shown herein. In decision-making and specifically to support targeting, one would opt to employ the proportion and the confidence interval for a given variable to rank all the administrative subunits (Giné-Garriga and Pérez-Foguet, under review-a), where top positions would denote highest priority. From Table 4, for instance, Rangwe could be easily identified as the most water poor division in Homa Bay (pi, access = 0.355; pi, time = 0.753), in which thus focus policy attention. Such prioritization, however, remains elusive where confidence intervals of the different subunits overlap. As general rule, it can be seen (Tables 4 to 7) that lower levels of confidence (α increases) give smaller confidence intervals, and hence reduced overlapping. Therefore, decision-makers would need to balance precision of final estimates (α) against robustness of statistics for planning purposes. For example, in Homa Bay and as regards time spent in water hauling (Table 4), one could target differently Riana (pi = 0.910) and Nyarongi (pi = 0.936) for confidence level of 80% (0.882– 0.932 and 0.913–0.953 respectively), though such discrimination would not be feasible with 90% confidence because of overlapping of the proportions with their corresponding intervals (0.875–0.938 and 0.906–0.957 respectively).

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711 Table 4 Estimated proportion and confidence interval of water-related indicators in Homa Bay (Kenya).

pi

α = 0,1

α = 0,2

pL,i - pU,i

pL,i - pU,i

Table 6 Estimated proportion and confidence interval of WASH indicators in Kibondo District (Tanzania). Use of Sanitation

Time to fetch watera

Access to improved waterpoints pi

α = 0,1

α = 0,2

pL,i - pU,i

pL,i - pU,i

707

Asego

0,510

0,449 - 0,569

0,462 - 0,556

0,807

0,755 - 0,851

0,766 - 0,842

Rangwe

0,355

0,298 - 0,414

0,310 - 0,401

0,753

0,696 - 0,802

0,708 - 0,792

Ward

pi

α = 0,1 pL,i -pU,i

Bunyambo

0,017

Busagara

0,049

Gwunumpu

Latrine Conditions

α = 0,2 pL,i -pU,i

pi

α = 0,1 pL,i - pU,i

α = 0,2 pL,i -pU,i

0,004 - 0,043

0,006 - 0,037

0,144

0,101 - 0,194

0,109 - 0,183

0,025 - 0,083

0,029 - 0,075

0,259

0,206 - 0,317

0,217 - 0,305

0,021

0,007 - 0,047

0,009 - 0,041

0,086

0,054 - 0,127

0,060 - 0,117

Ndhiwa

0,521

0,458 - 0,582

0,472 - 0,569

0,968

0,938 - 0,986

0,944 - 0,983

Nyarongi

0,663

0,615 - 0,708

0,625 - 0,699

0,936

0,906 - 0,957

0,913 - 0,953

Itaba

0,072

0,043 - 0,112

0,048 - 0,103

0,133

0,093 - 0,182

0,101 - 0,171

Riana

0,441

0,388 - 0,493

0,399 - 0,482

0,910

0,875 - 0,938

0,882 - 0,932

Kakonko

0,036

0,016 - 0,066

0,019 - 0,059

0,087

0,056 - 0,127

0,062 - 0,118

Kasanda

0,043

0,021 - 0,076

0,025 - 0,069

0,081

0,050 - 0,122

0,056 - 0,113

Kasuga

0,017

0,004 - 0,043

0,006 - 0,037

0,074

0,044 - 0,115

0,049 - 0,106

Kibondo Mjini

0,077

0,042 - 0,126

0,048 - 0,116

0,134

0,087 - 0,193

0,095 - 0,180

Kitahana

0,038

0,017 - 0,069

0,021 - 0,062

0,314

0,257 - 0,374

0,268 - 0,361

Kizazi

0,029

0,011 - 0,059

0,013 - 0,052

0,149

0,105 - 0,201

0,113 - 0,190

Kumsenga

0,013

0,002 - 0,039

0,003 - 0,033

0,057

0,029 - 0,096

0,034 - 0,087

Mabamba

0,017

0,004 - 0,042

0,006 - 0,036

0,139

0,098 - 0,188

0,106 - 0,177

Misezero

0,055

0,030 - 0,088

0,035 - 0,081

0,229

0,180 - 0,282

0,190 - 0,271

Mugunzu

0,000

0,000 - 0,016

0,000 - 0,012

0,098

0,063 - 0,143

0,070 - 0,133

Muhange

0,000

0,000 - 0,014

0,000 - 0,010

0,086

0,056 - 0,124

0,061 - 0,115

Murungu

0,035

0,015 - 0,068

0,018 - 0,061

0,097

0,061 - 0,143

0,068 - 0,133

Nyabibuye

0,028

0,011 - 0,057

0,013 - 0,050

0,094

0,061 - 0,138

0,067 - 0,128

Nyamtukuza

0,009

0,001 - 0,029

0,002 - 0,024

0,047

0,025 - 0,077

0,029 - 0,071

Rugenge

0,006

0,000 - 0,026

0,000 - 0,021

0,067

0,038 - 0,105

0,043 - 0,097

Rugongwe

0,029

0,012 - 0,055

0,015 - 0,049

0,206

0,160 - 0,257

0,169 - 0,246

Note: a) Households spending less than 30 min for one round-trip to collect water. In colour (red–orange–green), prioritization groups based on confidence intervals (α = 0.2).

One factor that challenges a reliable prioritization rank is the population variability in and within different administrative subunits. Depending on the nature of the indicator, a homogeneous pattern in the study area becomes more evident, primarily by i) reduced lengths of confidence intervals, and ii) greater overlapping of interval estimates. More specifically, the data from the three case studies show that indicators with low variability include gender disparities in the burden of collecting water and point-of-use water treatment; while at the other end of the spectrum, indicators presenting marked regional disparities are access to improved water supplies, and time spent in fetching water. It might be concluded from the data that the local scale of analysis do not add value where domestic habits and practices tend to a homogeneous behaviour, which would suggest that a regional estimate then may be enough for planning purposes. However, and prior to validating previous assumption, one would need to be sure that the indicator employed is adequate to describe patterns and trends in a given context. For example, in a peri-urban setting as in Manhiça, it can be seen that the improved/ unimproved classification of water supplies provides misleading information. Since the vast majority of households access an improved source, the level of service is better described through the availability of home connections (Table 7). As regards sanitation, it is gleaned from the estimates of Tables 5 and 6 that the approach adopted by the JMP does not lead to an obvious discrimination among administrative subunits in rural areas as Homa Bay and Kibondo, while an indicator related to open defecation status provides a clearer picture. In Manhiça, in contrast, the opposite applies (Table 7). In sum, where the studied variable shows a strong homogeneous pattern, the search for alternative indicators may be appropriate, since a larger sample size will probably prove ineffective to distinguish between different population groups. Despite the abovementioned restrictions, it is noteworthy that statistics shown in Tables 4 to 7 provide accurate inputs to policy-makers for the purpose of targeting. On the basis of performance level and therefore taking the estimates of proportions as reference points, for a given indicator one could group all subunits in different clusters, in such a way as to avoid overlap of their respective confidence intervals.

Note: In colour (red–orange–green), prioritization groups based on confidence intervals (α = 0.2).

For instance in Kibondo (Table 5), four different prioritization groups may be easily defined with regard to latrines' sanitary conditions, which ultimately allows a transparent identification of those subunits with poorest hygiene behaviour. Based on the same approach and to assess use of improved sanitation facility in Manhiça (Table 7), five target groups could be established. As regards the information provided by the waterpoint mapping, the analysis may focus on availability and geographic distribution of waterpoints. Without adequate combination of demographic data, however, this information might be misleading. Hence access indicators are usually assessed on the basis of standard assumption on the number of users per water source (i.e. the source:man ratio, which in Kenya stands at 250 people per public tap). Two different conclusions might be drawn from Table 8. First, it is observed that coverage levels of improved water points at the household (see Table 4) or at the waterpoint are substantially different; i.e. the standard source:man ratio is not followed up in practice. Second, access is dependent on the level of service; e.g. one out of five families in Homa Bay (21,1%) may get drinking water from an improved source, though this ratio is halved when water quality issues are taken into consideration.

Table 5 Estimated proportion and confidence interval of sanitation and hygiene indicators in Homa Bay (Kenya).

Use of improved facilities

Division pi Asego Rangwe Ndhiwa Nyarongi Riana

0,172 0,125 0,125 0,140 0,073

α = 0,1

α = 0,2

pL,i - pU,i

pL,i - pU,i

0,129 - 0,220 0,088 - 0,170 0,087 - 0,171 0,108 - 0,177 0,048 - 0,104

0,137 - 0,210 0,095 - 0,160 0,094 - 0,161 0,114 - 0,169 0,052 - 0,097

Open Defecation pi 0,358 0,430 0,531 0,630 0,667

Disposal of children stools

α = 0,1

α = 0,2

pL,i - pU,i

pL,i - pU,i

0,302 - 0,416 0,370 - 0,490 0,469 - 0,592 0,581 - 0,676 0,615 - 0,714

0,313 - 0,404 0,383 - 0,477 0,482 - 0,579 0,592 - 0,666 0,626 - 0,704

Note: In colour (red–orange–green), prioritization groups based on confidence intervals (α = 0.2).

pi 0,903 0,864 0,714 0,437 0,752

α = 0,1

α = 0,2

pL,i - pU,i

pL,i - pU,i

0,837 - 0,948 0,774 - 0,926 0,622 - 0,794 0,361 - 0,513 0,682 - 0,812

0,851 - 0,940 0,793 - 0,916 0,641 - 0,778 0,377 - 0,497 0,697 - 0,800

708

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

Table 7 Estimated proportion and confidence interval of WASH indicators in the Municipality of Manhiça (Mozambique).

Access to water (piped on premises)

Bairro

pi

α = 0,1

α = 0,2

pL,i - pU,i

pL,i - pU,i

Use of Sanitation

pi

Open Defecation

α = 0,1

α = 0,2

pL,i - pU,i

pL,i -pU,i

α = 0,1 pi

pL,i -pU,i

α = 0,2 pL,i -pU,i

Manhiça Sede

0,907

0,831 - 0,955

0,848 - 0,947

0,587

0,485 - 0,682

0,506 - 0,663

0,013

0,000 - 0,061

0,001 - 0,050

Tsá-Tsé

0,218

0,143 - 0,308

0,157 - 0,289

0,218

0,143 - 0,308

0,157 - 0,289

0,038

0,010 - 0,096

0,014 - 0,083

Mulembja

0,440

0,342 - 0,541

0,361 - 0,520

0,373

0,279 - 0,474

0,298 - 0,453

0,067

0,026 - 0,135

0,032 - 0,120

Ribangue

0,321

0,233 - 0,418

0,250 - 0,397

0,372

0,280 - 0,470

0,298 - 0,450

0,000

0 - 0,037

0 - 0,029

Balocuene

0,064

0,025 - 0,130

0,031 - 0,115

0,103

0,052 - 0,177

0,060 - 0,161

0,051

0,017 - 0,113

0,022 - 0,099

Timaquene

0,000

0 - 0,041

0 - 0,032

0,229

0,148 - 0,326

0,163 - 0,305

0,229

0,148 - 0,326

0,163 - 0,305

Chibucutso

0,000

0 - 0,039

0 - 0,030

0,080

0,035 - 0,151

0,042 - 0,136

0,067

0,026 - 0,135

0,032 - 0,120

Mitilene

0,000

0 - 0,039

0 - 0,030

0,067

0,026 - 0,135

0,032 - 0,120

0,347

0,255 - 0,447

0,273 - 0,426

Ribjene

0,000

0 - 0,039

0 - 0,030

0,013

0,000 - 0,061

0,001 - 0,050

0,613

0,511 - 0,707

0,532 - 0,688

Cambeve

0,286

0,202 - 0,382

0,218 - 0,361

0,208

0,134 - 0,298

0,148 - 0,279

0,026

0,004 - 0,079

0,006 - 0,067

Maciana

0,533

0,432 - 0,632

0,452 - 0,612

0,187

0,116 - 0,276

0,129 - 0,257

0,013

0,000 - 0,061

0,001 - 0,050

Maragra

0,987

0,938 - 0,999

0,949 - 0,998

0,880

0,799 - 0,935

0,817 - 0,926

0,000

0 - 0,039

0 - 0,030

Matadouro

0,573

0,471 - 0,670

0,492 - 0,650

0,333

0,243 - 0,433

0,261 - 0,412

0,000

0 - 0,039

0 - 0,030

Chibututuine

0,038

0,010 - 0,096

0,014 - 0,083

0,115

0,061 - 0,192

0,070 - 0,176

0,244

0,165 - 0,336

0,180 - 0,316

Wenela

0,960

0,899 - 0,989

0,913 - 0,985

0,440

0,342 - 0,541

0,361 - 0,520

0,013

0,000 - 0,061

0,001 - 0,050

Note: In colour (red – orange – green), prioritization groups based on confidence intervals (α = 0,2)

4.2. Communicating clear messages to policymakers The ultimate goal of sound sector-related data is to improve decisionmaking. To do this, two elements are necessary (Grosh, 1997): the data must be analyzed to produce outcomes that are relevant to the policy question, and the analysis must be disseminated and transmitted to policymakers. Unless data is easily accessible and is presented in a user-friendly format, decision makers will commonly do without the information. This section thus attempts to present a set of survey outputs to demonstrate that the approach adopted in this study produces pertinent sector data, which adequately exploited and disseminated might be employed by policy planners in decision-making processes. To begin, water and sanitation poverty maps are powerful instruments for displaying information and enable non-technical audiences to easily understand the context and related trends (Henninger and Snel, 2002). As observed from the tables discussed above, WASH-related poverty may follow a highly heterogeneous pattern, widely varying between

and within different administrative units; and mapping permits a feasible visualization of such heterogeneity (Davis, 2002). In addition, it provides a means for integrating data from different sources and from different disciplines (Henninger and Snel, 2002), which helps provide a complete picture of the context in which the service is delivered. In the end, mapping comes out an appropriate dissemination tool for sector planning, monitoring and evaluation support. The map in Fig. 1, for example, shows the spatial distribution of improved water sources in Homa Bay, and highlights the issues of functionality and seasonality. It is observed that the majority of audited sources were found operational and with no seasonality problems (71%), despite regional disparities. If such point-based information is combined with demographic data and the source:man ratio cited above, a coverage density map can be developed (Fig. 2) to show accessibility rather than availability aspects. It may be gleaned from the map that, on average, only 8.7% of population are properly served by functional and year-round improved waterpoints, i.e. the percentage of

Table 8 Water-related estimates in Homa Bay (Kenya), from the WPM. Division

Pop (2011)

No. imp WPs

Funct WPs

No. unim WPs

Unserved pop (by IWP)

IWPDa

FIWPDb

YR — FIWPDc

BS — FIWPDd

Asego Rangwe Ndhiwa Nyarongi Riana

94.950 109.148 59.211 56.912 64.673

33 57 24 48 25

31 50 21 38 18

No data No data 19 26 23

86.700 94.898 53.211 44.912 58.423

8.7% 13.1% 10.1% 21.1% 9.7%

8.2% 11.5% 8.9% 16.7% 7.0%

7.6% 9.8% 6.3% 13.6% 7.0%

5.8% 6.2% 5.9% 11.0% 4.3%

a b c d

IWPD: Improved waterpoint density; FIWPD: Functional improved waterpoint density; YRFIWPD: Year-round functional improved waterpoint density; BSFIWPD: Bacteriological safe functional improved waterpoint density.

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

Functional and year round Functional but seasonal Non-functional

Fig. 1. Distribution of functional improved water points, at location level.

population covered if it is assumed that each community tap only serves 250 people. Another concern in local decision-making is more related to the lack of transparent mechanisms to establish needs and priorities. Ideally, the most vulnerable segments of the population should be precisely targeted and then recipient of policy attention and public resources. And for this purpose, rankings and league tables are powerful instruments. To denote priorities and specifically to define prioritization criteria, two different approaches may be adopted. In terms of regional equity, the goal would be to reach a minimum coverage threshold in every administrative subunit. But based on an efficiency criterion, those subunits with highest number of potential beneficiaries should be first targeted, regardless of coverage. A combination of both criteria would also be feasible, despite resulting in a complex indicator.

< 5% 5 - 10% 10 - 15% 15 - 25% Equal or more than 25% No Funct Imp WP

Fig. 2. Density of year round and functional improved waterpoints, at location level.

709

As seen in Table 9, one different ranking is produced depending on each abovementioned criteria, showing both ranks poor correlation. For example, it is observed that Riana is prioritized as its open defecationindex stands at 67%, although in terms of potential beneficiaries, only roughly 60,000 people would beneficiate from the construction of new latrines. On the other hand, to defecate in the open is less common in Asego (36%), while beneficiaries from a hypothetical intervention would be raised up to 78.660. For planning purposes, the territorial equity criterion should be prioritized, as vulnerability is probably higher where coverage is lower (Jiménez and Pérez-Foguet, 2010). After targeting completion, each priority list could be easily related with specific remedial actions, therefore translating development challenges into beneficial development activities. Finally, few would dispute that pro-poor planning should be promoted to help address the issue of non-discrimination. The underlying hypothesis is that service level is highly dependent on social and economic conditions of population, which Table 10 confirms. From the data of two case studies (Kenya and Mozambique), and despite of poor control of confounding effects on variables measured, it is first observed that no significant differences exist with wealth regarding to access to improved water sources (pHB = 0.812 and pM = 0.14 respectively). A more in-depth analysis shows, however, that piped water on premises is enjoyed mainly by the wealthiest (pM = b0.001), while the “poor” significantly spend more time spent in hauling water (pHB = 0.016 and pM b 0.001). As regards sanitation, it is noted that use of improved latrines is positively related to wealth (pHB b 0.001 and pM = b 0.001). And for instance, the richest 25% of the population is almost ten (Homa Bay) or six (Manhiça) times as likely to use an improved sanitation facility as the poorest quartile, while the poorest 25% is two/ twenty times more likely to practise open defecation than the richest quartile. Much like water supply and sanitation, considerable differences are also found regarding to hygiene practices between the rich and the poor. The percentage of households with adequate point-of-use water treatment increases with wealth status (pHB = 0.043 and pM = b 0.001). And socio-economic status of the household shows strong association with safety in disposal of children's faeces (pHB = 0.028 and pM = 0.007). 5. Conclusions The delivery of water and sanitation services together with the promotion of hygiene is central to public health. In recent years, service delivery has shifted to decentralised approaches; on the basis that decentralisation will favour local needs and priorities. Any prospect to develop more pro-poor policies, though, depends upon real efforts to strengthen the capacity of decentralised authorities. Integral to this challenging process, and to enable policymakers to move from opaque to informed discussions, accurate and reliable data at local level have to be accessible, i.e. routinely collected and adequately disseminated. Against this background, the aim of this article is to develop a cost-effective method for primary data collection which ultimately produces estimates accurate enough to feed into decision-making processes. First, a simplified survey design for WASH data collection is presented, which improves on other existing methodologies in various ways. The approach adopted combines data from two different information sources: the water point and the household. It takes the WPM as starting point, as a method with increasing acceptance amongst governments and practitioners to inform the planning of investments when improving water supply coverage. Since mapping entails as part of the survey design specifications complete geographic representation of the study area, a stratified household-based survey is undertaken in parallel, in which a sample of households is selected from each stratum. In so doing, the risk of homogeneity within the strata remains relatively low, thus enabling reduced “design effects” in sample size determination.

710

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711

Table 9 Priority ranks for access to sanitation, based on open defecation practice, in Homa Bay (Kenya). Division

Population 2011

pi

Rank A (equity)

pL,i–pU,i α = 0.2

Rank A' (equity)

No served population

Rank B (efficiency)

Riana Rangwe Ndhiwa Nyarongi Asego

64.673 56.912 59.211 109.148 94.950

0.667 0.630 0.531 0.430 0.358

1 2 3 4 5

0.626–0.704 0.592–0.666 0.482–0.579 0.383–0.477 0.313–0.404

1 1 2 3 4

59.965 48.944 51.810 95.505 78.660

3 5 4 1 2

Second, the analysis of data in the discussion has produced valuable outputs that might be further exploited for local level policymaking support. It has shown how data can feed into planning and targeting decisions in a range of different ways. However, to offer relevant guidance to the policy question, the analysis must be disseminated effectively. Maps or simple thematic indices are adequate tools to capture the attention of policymakers, as they transmit a clear picture easily and accurately. Nevertheless, it is noteworthy that specific challenges remain elusive to effectively improve decentralised planning, primarily the continued use of the collected data in decision-making, and the development of appropriate updating mechanisms (WaterAid, 2010). In the short term, the effective exploitation of planning data by local decision-makers demands continued support from multi-stakeholder alliances between governments, NGOs, academics and consultants. In the medium term, however, political will and commitment at all levels, i.e. from central government to local authorities, is imperative to ensure that improved use of data results in effective pro-poor planning. Similarly, the evaluation framework needs to be rethought from the viewpoint of sustainability, so that it could be updated autonomously by local stakeholders or replicated elsewhere. In this regard, a major shortcoming is the trade-off between the scope and quality of the data required for decision-making support and the complexity of updating mechanisms (WaterAid, 2010). These two challenges may suggest the way forward.

Acknowledgements The authors would like to extend thanks to all families who participated in the study. Further thanks go to ONGAWA and to Kibondo District Water Department for their support to undertake the survey in Kibondo District, in Tanzania; to UNICEF (Kenya Country Office) and to Homa Bay District Water Office and District Public Health Office, in Kenya; and to UN Habitat (Country Office) and the Municipality of Manhiça, in Mozambique. This study has been partially funded by the Centre de Cooperació per al Desenvolupament (Universitat Politècnica de Catalunya) and the Agencia Española de Cooperación Internacional para el Desarrollo [reference 11-CAP2-1562]. Financial support of the Agencia General d´Ajuts Universitaris i de Recerca -AGAUR- (Generalitat

Table 10 Access to water, sanitation and hygiene by wealth status. Indicator

Access to improved water sources One way distance to water source (km) Access to and use of improved sanitation Point-of-use water treatment Disposal of children's stools

Pearson Chi-square exact sig. (2-sided)a Homa Bay (Kenya)

Manhiça (Mozambique)

p p p p p

p p p p p

= 0.812 = 0.016 b 0.001 = 0.043 = 0.028

= 0.140 b 0.001 b 0.001 b 0.001 = 0.007

a In Pearson's chi-square test, the null hypothesis is independence, and the value p = 0.05 is used as the cut-off for rejection or acceptance.

de Catalunya) through the Beatriu de Pinos grant (BP-DGR 2011) to Dr A. Jiménez is also acknowledged. References Bartram J, Corrales L, Davison A, Deere D, Drury D, Gordon B, et al. Water safety plan manual: step-by-step risk management for drinking-water suppliers. Geneva: World Health Organization; 2009. Bennett S, Woods T, Liyanage WM, Smith DL. A simplified general method for clustersample surveys of health in developing countries. World Health Stat Q 1991;44: 98–106. Booysen F, van der Berg S, Burger R, Maltitz Mv, Rand Gd. Using an asset index to assess trends in poverty in seven sub-Saharan African countries. World Dev 2008;36: 1113–30. Cairncross S, Hunt C, Boisson S, Bostoen K, Curtis V, Fung IC, et al. Water, sanitation and hygiene for the prevention of diarrhoea. Int J Epidemiol 2010;39(Suppl. 1):i193–205. Cochran WG. Sampling Techniques. 3rd ed. New York: John Wiley and Sons; 1977. Crook RC. Decentralisation and poverty reduction in Africa: the politics of local-central relations. Public Adm Dev 2003;23:77–88. Curtis V, Cairncross S. Effect of washing hands with soap on diarrhoea risk in the community: a systematic review. Lancet Infect Dis 2003;3:275–81. Davis B. Is it possible to avoid a lemon? Reflections on choosing a poverty mapping method. Rome: Food and Agricultural Organization of the United Nations; 2002. Devas N, Grant U. Local government decision-making — citizen participation and local accountability: some evidence from Kenya and Uganda. Public Adm Dev 2003;23: 307–16. Esrey SA, Potash JB, Roberts L, Shiff C. Effects of improved water supply and sanitation on ascariasis, diarrhoea, dracunculiasis, hookworm infection, schistosomiasis, and trachoma. Bull World Health Organ 1991;69:609–21. Feachem RG. Interventions for the control of diarrhoeal diseases among young children: promotion of personal and domestic hygiene. Bull World Health Organ 1984;62: 467–76. Fewtrell L, Kaufmann RB, Kay D, Enanoria W, Haller L, Colford JJM. Water, sanitation, and hygiene interventions to reduce diarrhoea in less developed countries: a systematic review and meta-analysis. Lancet Infect Dis 2005;5:42–52. Filmer D, Pritchett LH. Estimating wealth effects without expenditure data-or tears: an application to educational enrollments in states of India. Demography 2001;38: 115–32. Frerichs RR, Tar KT. Computer-assisted rapid surveys in developing countries. Public Health Rep 1989;104:14–23. Giné-Garriga R, Pérez-Foguet A. Sample size determination for local household-based surveys in water, sanitation and hygiene sector. Ecol Indic 2013a. [under review]. Giné-Garriga R, Pérez-Foguet A. Water, sanitation, hygiene and rural poverty: issues of sector monitoring and the role of aggregated indicators. Water Policy 2013b. in press. Giné-Garriga R, Jiménez A, Pérez-Foguet A. A closer look at the sanitation ladder: issues of monitoring the sector. 35th WEDC International Conference. Water, Engineering and Development Centre. Loughborough, UK: Loughborough University; 2011. Government of Liberia. Liberia Waterpoint Atlas. Monrovia: Ministry of Public Works; 2011. Government of Sierra Leone. Sierra Leone Waterpoint Report. Freetown: Ministry of Energy and Water Resources; 2012. Grosh ME. The policymaking uses of multitopic household survey data: a primer. World Bank Res Obs 1997;12:137–60. Henninger N, Snel M. Where are the poor? Experiences with the development and use of poverty maps. Washington, D.C.: World Resources Institute; 2002 Howard G, Ince M, Smith M. Rapid assessment of drinking-water quality: a handbook for implementation. Geneva: World Health Organization and United Nations Children Fund; 2003 [draft]. Hunt C. How safe is safe? A concise review of the health impacts of water supply, sanitation and hygiene. . WELL Study No. 509. 509London and Loughborough: WELL Resource Centre (WEDC & LSHTM); 2001. Jiménez A, Pérez-Foguet A. Building the role of local government authorities towards the achievement of the human right to water in rural Tanzania. Nat Res Forum 2010;34:93–105. Jiménez A, Perez-Foguet A. Water point mapping for the analysis of rural water supply plans: case study from Tanzania. J Water Resour Plann Manag-ACSE 2011;137:439–47. Jiménez A, Pérez-Foguet A. Implementing pro-poor policies in a decentralized context: the case of the Rural Water Supply and Sanitation Program in Tanzania. Sustain Sci 2011;6:37–49.

R. Giné-Garriga et al. / Science of the Total Environment 463–464 (2013) 700–711 Jiménez A, Pérez-Foguet A. Quality and year-round availability of water delivered by improved water points in rural Tanzania: effects on coverage. Water Policy 2012;14: 509–23. Jiménez A, Molinero J, Pérez-Foguet A. Monitoring Water Poverty: A Vision from Development Practitioners. In: Llamas MR, LMC, Mukherji A, editors. Water Ethics: Marcelino Botin Water Forum 2007. London: Taylor & Francis; 2009. Joint Monitoring Programme. Global Water Supply and Sanitation Assessment 2000 Report Joint Monitoring Programme for Water Supply and Sanitation. Geneva/New York: WHO/UNICEF; 2000. Joint Monitoring Programme. Core questions on drinking-water and sanitation for household surveys. Geneva/New York: WHO/UNICEF; 2006. Joint Monitoring Programme. Report of the JMP Technical Task Force Meeting Monitoring of drinking-water quality. Geneva/New York: WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP); 2010. Joint Monitoring Programme. Report of the First Consultation on Post-2015 Monitoring of Drinking-Water and Sanitation. Berlin: WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP); 2011. Joint Monitoring Programme. Progress on Drinking Water and Sanitation: 2012 Update. Joint Monitoring Programme for Water Supply and Sanitation. Geneva/New York: WHO/UNICEF; 2012a. Joint Monitoring Programme. Report of a Technical Consultation on the Measurability of Global WASH Indicators for Post-2015 Monitoring. New York: WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP); 2012b. Joint Monitoring Programme. Report of the Second Consultation on Post-2015 Monitoring of Drinking-Water, Sanitation and Hygiene. The Hague: WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP); 2012c. Kish L. Design and Estimation for Domains. Statistician 1980;29:209–22. Leemis LM, Trivedi KS. A comparison of approximate interval estimators for the Bernoulli parameter. Am Stat 1996;50:63–8. Lemeshow S, Stroh G. Sampling Techniques for Evaluating Health Parameters in Developing Countries. Washington, D.C.: National Research Council; 1988. Lwanga SK, Lemeshow S. Sample size determination in health studies: a practical manual. Geneva: World Health Organization; 1991.

711

Macro International Inc.. Sampling Manual. Calverton, Maryland: Demographic and Health Surveys - III; 1996. Manun'Ebo M, Cousens S, Haggerty P, Kalengaie M, Ashworth A, Kirkwood B. Measuring hygiene practices: a comparison of questionnaires with direct observations in rural Zaire. Trop Med Int Health 1997;2:1015–21. O'Donnell O, Doorslaer Ev, Wagstaff A, Lindelow M. Analyzing health equity using household survey data: a guide to techniques and their implementation. Washington, D.C.: The World Bank; 2007. Pearce J, Howman C. RWSN Water Point Mapping Group: A Synthesis of Experiences and Lessons discussed in 2012. St. Gallen: Rural Water Supply Network; 2012. Reiczigel J. Confidence intervals for the binomial parameter: some new considerations. Stat Med 2003;22:611–21. Steiner S. Decentralisation and poverty: conceptual framework and application to Uganda. Public Adm Dev 2007;27:175–85. Stoupy O, Sugden S. Halving the Number of People without Access to Safe Water by 2015 — A Malawian Perspective. Part 2: New indicators for the millennium development goal. London: WaterAid; 2003. Sutton S. The risks of a technology-based MDG indicator for rural water supply. 33rd WEDC International Conference. Water, Engineering and Development Centre. Accra, Ghana: Loughborough University; 2008. United Nations Children's Fund. Multiple Indicator Cluster Survey Manual 2005. New York: UNICEF, Division of Policy and Planning; 2006. United Nations Children's Fund. A WASH Baseline Survey under the NL–UNICEF partnership for WASH. One Million Initiative (Mozambique). Maputo: UNICEF and WE Consultant; 2009. WaterAid. Water point mapping in East Africa. Based on a strategic review of Ethiopia, Tanzania, Kenya and Uganda. London: WaterAid; 2010. WaterAid, ODI. Learning for Advocacy and Good Practice — WaterAid Water Point Mapping. London: Overseas Development Institute; 2005. World Health Organization. GLAAS 2012 Report: The Challenge of Extending and Sustaining Services. UN-Water Global Analysis and Assessment of Sanitation and Drinking-Water (GLAAS). Geneva: World Health Organization; 2012.