THE MPA IN ACTION4.5 Selecting the communitiesWhen used for monitoring or self-evaluation, theMPA helps its practitioners assess services alreadyoperating. When used for planning of newprojects, the assessments will establish how existingsystems have been established <strong>and</strong> are managed.MPA is most appropriate to use in the followingsituations:●●●●The water <strong>and</strong> sanitation systems that will beassessed should have been functioning forsome time, at least 2-3 years or longer.If they are still functioning, various aspects oftheir functioning will be assessed. If notfunctioning, reasons leading to the system’sfailure will be investigated <strong>and</strong> documented.The service should have some form ofdecentralized management, i.e., it should notbe exclusively managed by an external agency.If the services are entirely externally managed,tools in addition to MPA will need to be used.The project organization <strong>and</strong> communityshould further be willing to participate in theassessments. 12In large programs, the assessments will usuallybe carried out with a r<strong>and</strong>om sample ofcommunities. A r<strong>and</strong>om sample gives everycommunity in a project or program the samechance to be included in the assessment. In ar<strong>and</strong>om sample, each community has the sameprobability of being approached to participatein an assessment. Each potential community isalso selected independently from others. Thoseeligible would be all communities in a particularprogram area when the program makes nodistinction for size <strong>and</strong> type of communities thatmay participate. Usually, however, programs havespecific selection criteria for communities. The“sampling pool” (known in statistics as theuniverse) will then consist of all communities thatmeet the selection criteria <strong>and</strong> have water supply<strong>and</strong>/or improved sanitation facilities installedunder the ongoing or completed program.Alternatively, when MPA is used as a projectplanning or baseline building method, theselected communities may have older serviceprovisions, either made by themselves or underearlier programs or both. The incoming programwill be aimed at enabling these communities toupdate, exp<strong>and</strong>, improve, or replace the systems.In a non-r<strong>and</strong>om sample, the findings <strong>and</strong>conclusions from selected communities only holdfor the specifically selected cases, <strong>and</strong> are notapplicable for all communities in the program(s)from which they were selected.Special caution is required if organizations orcommunities take part in comparative studiesbased on an open invitation from the organizers.Chances are that in such cases of self-selection,the sample will have a bias towards moreparticipatory, socially inclusive, <strong>and</strong> gender-awarecommunities. In such situations, the variation inapproaches <strong>and</strong> results between individualcommunities may still be large enough to bringout differences that are significant, that is, to bedue to more than pure chance. However, in askewed sample, e.g., when selection procedureshave led to an overrepresentation of more12 In keeping with participatory research principles, evaluation exercises may not be imposed on unwilling ordisinterested communities. However, working only with eager <strong>and</strong> willing communities can seriously bias thesample <strong>and</strong> results of assessments. It is therefore important to develop a rigorous sampling plan including analternative reserve list with additional communities, numbering at 20 percent of the required sample. Samplecommunities from the primary list are first contacted as planned. In case resistance or lack of interest inparticipating is encountered in a community, a genuine attempt should first be made to ascertain reasons bycontacting secondary sources of information <strong>and</strong> key individuals. If by doing so trust can be established, MPAassessments can be carried out as planned. If not, the case should be documented <strong>and</strong> another comparablecommunity selected from the reserve list. Such documentation usually provides valuable learning about pastproject approaches <strong>and</strong> their consequences.34
THE MPA IN ACTIONsuccessful or unsuccessful projects, regressionequations can not be applied to help predict whichinputs lead to better sustainability <strong>and</strong> use ofservices.4.6 Sampling proceduresThe sample size <strong>and</strong> rigor of sampling procedureswill vary with the purpose for which the MPA isbeing used, e.g., as a part of training formanaging water-sanitation projects, for casestudies, program planning, monitoring orevaluation, or research purposes, <strong>and</strong> dependingon whether a statistical analysis is required.For most evaluation research <strong>and</strong> monitoringpurposes, the number of communities in thesample would vary with the size of the project <strong>and</strong>the types of interventions it makes. The aim is toinvolve enough communities to provide a goodcross-section of the typical technical, social,economic, cultural, political, administrative, <strong>and</strong>environmental conditions in the project area,without confounding biases in selection.If overall conditions are homogenous, a simpler<strong>and</strong>om sample can be drawn. To do so, eachcommunity in the program(s) being sampled getsits own number, beginning with number one. Eachnumber is written on a separate slip of paperwhich is tightly folded. The slips are placed in abox or bag. They are stirred well until all havemixed. The required number of communities forthe sample (see below under sample size) is thendrawn from the container.The alternative is to use a table with r<strong>and</strong>omnumbers. Assuming that a sample of 40communities needs to be drawn for a programwith 600 communities, a numbered list ofnames is prepared from 1 to 600. To drawthe sample, one can start reading numbersin the table that consist of any three figures.The starting point may be anywhere in thetable. The first 40 numbers thus generatedbetween 1 <strong>and</strong> 600 are the numbers of thecommunities that will be contacted.Continuing until one has 50 communities thusidentified will provide a reserve in case of laterrefusals to participate.When conditions are heterogeneous, communitiesare grouped (stratified) according to their maincharacteristics. They are then selected at r<strong>and</strong>omin numbers proportional to their presence in theprogram. Box 16 lists some factors that may be ofBox 16 Possible factors for consideration in decisions on type of sampling● Environmental <strong>and</strong> technical conditions: type of water sources (ground <strong>and</strong> surface water),availability <strong>and</strong> quality of fresh water, water resources, developments in water <strong>and</strong> l<strong>and</strong> use,technologies used for water supply <strong>and</strong> sanitation.● Demographic conditions <strong>and</strong> developments: population size, density, growth, <strong>and</strong> migration.● Economic conditions: economic base (e.g. subsistence, cash crop, or industrial or services economy),communications (near major cities, well-connected or isolated), character of the settlement (ruralvillage, small market town, or low-income urban), economic growth.● Socio-cultural conditions: religious, caste <strong>and</strong> ethnic composition, male <strong>and</strong> female literacy levels,heterogeneous or homogeneous society, seclusion of women, etc.● Political <strong>and</strong> administrative conditions: decentralization <strong>and</strong> devolution, types <strong>and</strong> legal status ofwater <strong>and</strong> sanitation management organizations.(See example of sample selection criteria in the Lao PDR case study, Section 2 of this report.)35